Party Animals? Extreme Partisan Polarization and Dehumanization


James L. Martherus, Vanderbilt University

Andres G. Martinez, Sonoma State University

Paul K. Piff, University of California, Irvine

Alexander G. Theodoridis, University of California, Merced


The visceral, affectively-charged, identity-based, and often negative nature of partisan polarization in the United States has been the subject of much scholarly attention. A burgeoning literature on partisan dehumanization suggests a qualitative shift may be occurring—from partisan animosity to partisan dehumanization (the denial of human characteristics to out-partisans). Recent examples of dehumanizing rhetoric by elites on both sides abound: Bill Maher called Republicans “treasonous rats.” Alex Jones described Democrats as “the ultimate cowardly sacks of garbage,” Harry Reid dubbed President Trump the GOP’s “Frankenstein monster,” and Eric Trump said Democrats investigating his father were “not even people.”

In light of new work in social-personality psychology, we investigate the extent to which contemporary hyper-polarization of the electorate has devolved into a willingness by voters to apply dehumanizing metaphors to out-partisans. To do this, we bring to bear three novel large-sample, broadly representative online surveys, fielded over the course of four years, and across two presidential administrations.

We begin by looking at two different measures of dehumanization (one subtle and one blatant). This uncovers striking, consistent observational evidence that MOST partisans dehumanize members of the opposing party.

We delve into the relationships between dehumanization and important hallmarks of partisan intensity and polarization. We find that inter-partisan dehumanization is most closely related to extreme affective polarization. In addition, we show that dehumanization is associated with biased partisan-motivated reasoning and correlated with the degree to which partisans hold authoritarian/fixed worldviews.

We uncover one possible source of this troubling phenomenon with an experiment offering causal leverage to examine openness to dehumanization in the processing of new information about misdeeds by in- and out-partisans. Participants were exposed to identical information about a melee at a gathering, with the partisanship of the perpetrators randomly assigned. We find pronounced willingness by both Democrats and Republicans to differentially dehumanize members of the out-party (even when the transgression in question was the same).

These findings illuminate the nature, depth, and form of contemporary partisan polarization. Inter-partisan dehumanization may (in part) explain the increased reluctance to find common ground on political issues, and hence constitute a growing threat to the democratic enterprise. And, if not curbed, tendencies toward dehumanization may presage (as they have in other contexts) a heightening of partisan violence.


Position Taking on the President’s Agenda


Jason S. Byers, University of Georgia

Laine P. Shay, University of Utah

Do elections influence a member of Congress’s decision to engage in position-taking via casting a vote on a legislative roll call? Several scholars suggest that an increase in electoral vulnerability for a legislator should be associated with a decrease in abstention behavior in congressional roll call voting.  However, there are mixed empirical findings to support this relationship. We argue that one factor previous studies have not accounted for is the salience of the roll call vote.

We suggest that electorally vulnerable members of Congress are more likely to cast a vote on a roll call that has been publicly addressed by the president. Roll call votes that the president has publicly addressed have been shown to be more salient than those that have not received attention from the president. In particular, previous studies have found that this type of roll call vote can influence a member of Congress’s electoral prospects. Thus, it might be politically unwise for them to skip this type of salient vote. Additionally, casting a vote on these salient roll calls might help an electorally vulnerable member fend off an attack from a political challenger highlighting how the incumbent has shirked his or her legislative responsibilities. Therefore, we expect that electorally vulnerable members are less likely to abstain on roll call votes that the president has publicly addressed, compared to their electorally secure counterparts in the House of Representatives.

To test this claim, we examine all roll call votes from legislators serving in the House of Representatives between the 84th and the 112th Congress. We classify the roll call votes by two dimensions: 1) whether the roll call vote was close or lopsided in its final outcome; 2) whether the president offered a position on the roll call. We identify roll call votes that the president has publicly addressed via CQ Presidential Support Scores. Our dependent variable is the percentage of abstentions by a member in a given Congress within a classification scheme. We consider an abstention one in which a member does not cast a yea, nay, or paired vote. Using this measure, a member can potentially receive a value between zero (indicating they did not abstain on any roll call vote within a classification scheme and in a Congress) and one (indicating they abstained on all roll call votes within a classification scheme and in a Congress).

To examine the relationship between electoral vulnerability and abstentions, we estimate the coefficients with a fractional logit model. We find that for votes that the president has publicly addressed, electorally vulnerable members of Congress are less likely to abstain from voting compared to their electorally secure counterparts, regardless if the vote is lopsided or close in its final outcome. For lopsided votes that the president has publicly addressed, we find that the most electorally secure member abstains on 20 percent more roll call votes than the most electorally vulnerable member. Similarly, for close votes that the president has publicly addressed, we find that the most electorally secure member abstains on 33 percent more votes than the most electorally vulnerable member. We also find that electorally vulnerable members of Congress are less likely to abstain on close votes that the president has not addressed.  Finally, we find that a member’s electoral vulnerability has limited influence on their abstention behavior on lopsided votes that the president has not publicly addressed.  Overall, our findings suggest that electorally vulnerable members are more likely to cast a vote on a salient roll call.

Our results provide many important implications regarding legislative behavior. First, our results indicate that the type of vote matters when examining voting behavior. Salient roll calls are more likely to attract electorally vulnerable members to engage in position taking by casting a vote. Most students of legislative politics tend to aggregate roll calls when examining congressional voting behavior.  Our findings suggest that it is important for scholars to sometimes consider disaggregate roll call votes in order to uncover the precise dynamics of legislative behavior.  Second, these findings provide additional insight into the relationship between the president and members of Congress. These findings suggest that the president is a central focus in a legislator’s electoral strategy.  Lastly, previous studies find limited empirical evidence of electoral vulnerability influencing a member’s abstention behavior.  This study provides a bridge for the mixed findings regarding the relationship between competitive elections and congressional abstention.

In-Group Love Versus Out-Group Hate: Which Is More Important to Partisans and When?



Karyn Amira, College of Charleston

Jennifer Cole Wright, College of Charleston

Daniela Goya-Tocchetto, Duke University

Affective polarization is on the rise, driving historically high levels of negative feelings towards out party members. When political intergroup conflict occurs, what is the primary reaction that people have: do they run to defend their in-group, or do they lash out against the out-group? Are there conditions under which this decision changes? We test these questions in a series of 4 studies.

In our first study we gave partisans a forced choice: choose to promote an online news article that lauds their own party or one that denigrates the opposing party. Though others have shown that affective polarization leads to in-party help rather than out-party harm, our design tests which bias is primary in a non-zero sum way: either favoring the in-group (without thereby harming the out-group) or harming the out-group (without thereby favoring the in-group). Consistent with previous results, we find evidence that subjects prefer to help their own party’s reputation rather than harm the opposing party (67% to 33% of subjects, respectively).

While this is good news, we also show that this more benign form of prejudice can shift focus under situations of symbolic threat—activating out-group animosity. In a follow up study, new subjects made the same forced choice: promote the news article that helps their own party, or one that harms the opposing party. However, in this study some subjects randomly assigned to a treatment had to read another story that criticized the morality of their own partisan group (a symbolic threat) before making their forced choice. Those in this “threat condition” were more likely to shift towards punishing the opposing party compared to the control group. The figure below (from Study 2) shows the shift between people who saw the symbolic moral threat before selecting (1) vs. those who did not (0).


When we added the option to do nothing (Study 3), the moral threat still caused subjects to harm the out-party, but it also caused them to opt out entirely. In Study 4, harming the out-party became even more frequent when the targets of the moral threat were ordinary party members in the mass public (rather than elites in Washington), which could make the threat feel more personal. In Study 4, very few subjects opted out of choosing one of the two articles (noticeably more than in Study 3)—showing a higher willingness to engage in out-group denigration.

While it does not seem feasible to completely suppress our natural tribal tendencies to create group identities, fostering a positive attitude towards in-group members is more conducive to social cooperation than the denigration of the out-group. This positive tendency serves as a straightforward way to reaffirm those group identities. To that extent, showing that protection of the partisan in-group is indeed a core psychological trait is an encouraging result in an age in which polarization seems to be bringing out the worst in people.

On a more worrisome note, studies 2-4 found that a symbolic moral threat to one party’s underlying core values caused respondents to lash out at the opposing party. This response was expected, based on the literature claiming that one’s identification with a political party resembles other types of group identities (e.g., religion, sports) that constitute a significant part of one’s personal identity.

Although we are able to replicate reassuring evidence that partisans do not go out of their way to harm the opposing party, our findings also suggest that more outward, aggressive behavior could be easily triggered by threats perceived as moral. This is of especial importance in a time when articles such as “The Moral and Intellectual Bankruptcy of the Republican Party” (The Guardian), “The Deeply Immoral Values of Today’s Republican Leaders” (Huffington Post), and “The GOP’s Moral Rot is the Problem, not Donald Trump Jr.” (Washington Post) are becoming commonplace.


Not Dead Yet: Political Learning from Newspapers in a Changing Media Landscape

Erik Peterson, Texas A&M University

Local media outlets are a crucial source of information about politics, but today they face several challenges. The vast expansion in media choice brought about by the introduction of cable television and the internet means local newspapers and television stations now reach a smaller audience than in the past. Widespread staff cuts have also left these outlets with a more limited capacity to produce political news. To what extent can local media still inform the public in this changing landscape?

I take up this question by studying newspapers’ ability to inform the public about their member of Congress. This is particularly important to understand as local newspapers are the primary source of original reporting on this topic. The national media devotes limited attention to the actions of individual representatives and local television stations often rely on newspaper coverage rather than producing their own. Newspapers also exemplify the broader set of challenges facing local media, as the industry has experienced rapid declines in readership and staffing over the past decade.

To examine the effects of local newspapers, I create new measures of the fit between newspaper markets and congressional districts using the geographic distribution of contemporary newspaper sales. This measure of “congruence,” originally introduced in a 2010 study of local newspapers by James M. Snyder and David Stromberg, captures the amount of coverage about their representative in the U.S. House that public can find from the newspapers available in their area, with higher levels of congruence indicating greater coverage availability.

To measure the consequences of the newspaper environment, I pair this with the Cooperative Congressional Election Study, a series of large-scale public opinion surveys conducted from 2006-2017 that include several questions assessing the public’s awareness of their member of Congress. Finally, to isolate the influence of newspapers from other factors that contribute to how much the public knows about their representative, I hold constant election-related and politician-specific variables with a research design that only uses differences in congruence that occur within a congressional district in a given year.

Using this approach, I find that the local newspaper environment continues to influence political awareness despite these challenges. For instance, a typical shift in congruence — equivalent to roughly 18 additional stories about an individual’s member of Congress appearing in newspapers during a term — produces an expected 1.4 percentage point increase in the public’s ability to assess their representative’s ideological record. A supplementary analysis indicates that even those with a limited interest in politics still benefit from the availability of newspaper coverage.

However, despite their continued influence, the broader crisis facing local newspapers has taken a toll on their ability to inform the public. In comparison to Snyder and Stromberg’s earlier study, which used a similar approach with data spanning 1982-2004, the contemporary effects of newspapers on political awareness have decreased and are now one-half to one-third their former size.

My study illustrates the importance of newspapers, particularly their continued ability to inform the public despite adverse shifts in the media landscape. At the same time, the comparison with the past shows just how much the local media’s influence has eroded. This complements a growing body of recent research that explores the political impact of changes in the local media environment (Martin and McCrain 2019, Darr, Hitt and Dunaway 2018, Hayes and Lawless 2018). Given that the challenges faced by newspapers in the United States are similar to those facing legacy media outlets around the world, this also highlights the need for more research that reconsiders theory and evidence about these outlets’ role in informing the public.

Opinion Shift and Stability: The Information Environment and Long-Lasting Opposition to Trump’s Muslim Ban


Kassra A. R. Oskooii, University of Delaware

Nazita Lajevardi, Michigan State University

Loren Collingwood, University of California–Riverside

A few days after the tragic San Bernardino attacks, presidential hopeful Donald Trump first promised a “Muslim Ban.” Less than one week into his presidency, Trump made good on his promise and signed executive order (EO) 13769, which denied citizens of seven Muslim-majority countries entry into the United States. In the next two days, as immigration lawyers at airports around the country filed writs of habeas corpus for the release of individuals affected by the ban, impromptu mass demonstrations involving tens of thousands of protesters erupted in cities and airports nationwide. These protests received widespread media coverage and elicited criticism from media pundits and politicians on both sides of the aisle.

In previous work published in Political Behavior, we demonstrated that individual-level opinions moved decisively against the ban in a matter of days. We argued that an influx of information depicted the ban as distinctly un-American and in violation of the principle of American religious freedom, which invoked meaningful and new evaluative criteria about the policy. This, in turn, provoked attitude change, particularly among high American identifiers, who likely saw a clear incompatibility between American values and the EO.

The present study extends previous research by examining whether within-subject shifts against the ban lasted over an extended period of time. Inquiring into the stability of rapid changes in attitudes is important because if preferences regress back to the mean, the impact of political communication on opinions may be just temporary. But, if opinions remain relatively stable once altered, fairly significant and one-sided changes in the information environment could have longstanding and substantively meaningful impacts on preferences.

Evidence from a large corpus of transcripts (cable news broadcasts and four national newspapers) demonstrates that the ban not only remained on the agenda, but that political communication between February 2017 and January 2018 was highly critical of it and did not present a significant counter-narrative in favor of the EO. As Figure 1 helps illustrate, critical coverage of the ban (blue line) across the entire time frame dominated positive or pro-ban coverage (red line).

Figure 1: Valence Coverage of Newspaper Articles about the “Muslim ban/Travel ban” Over Time. Source: New York Times, Washington Post, Wall Street Journal, and USA TodayOskooii2.2

Relying on a timely, three-wave panel dataset, we then examined competing hypotheses about attitude stability and change, and found that individual-level opinions remained largely stable once they had shifted against the ban (See Table 1). We also found that high American identifiers, who shifted against the ban in the days after it was announced, remained less supportive one year later. As Figure 2 illustrates, high American identifiers’ ban attitudes in T2 and T3 are nearly identical, and less supportive of the travel ban than in T1.


Table 1: Difference of Means T-Tests of Ban Attitudes between T1 (Jan. 24-27, 2017), T2 (Feb. 2-8, 2017), and T3 (Jan. 8-14, 2018).Oskooii3.2

Our study contributes to research on public opinion and on race and ethnic politics. It suggests that mass movements that successfully prime American inclusiveness can durably move individuals, especially high American identifiers, against new policies that target racial, ethnic, or religious minority groups. Given that American identity is typically associated with anti-immigrant policy preferences (Citrin, Reingold and Green, 1990; Frendreis and Tatalovich, 1997; Schildkraut, 2003) and that group-centric attitudes tend to be highly crystallized and fixed (Kinder and Sanders, 1996; Nelson and Kinder, 1996), rapid and durable within-subject opposition toward the “Muslim ban” is noteworthy. It indicates that policy attitudes related to highly disliked groups, such as Muslims, are perhaps more malleable than previously assumed. Additionally, the findings offer some hope that some citizens are still open to considering new evaluative criteria that may challenge their priors despite the state of partisan polarization and citizens’ tendency to support their existing evaluations even in the face of disconfirming information.

Figure 2: Strength of American Identity and Attitudes Toward the Muslim BanOskooii4.2.png



Political Considerations in Nonpolitical Decisions: A Conjoint Analysis of Roommate Choice

Richard M. Shafranek

Northwestern University

Partisan political considerations sometimes influence decisions and interactions outside the context of politics. Scholars commonly attribute the influence of political preferences on social decisions to dislike for members of the other party. Studies have demonstrated that partisans may discriminate against their political opponents in circumstances ranging from academic hiring to everyday economic transactions – observations that have been echoed even in media outlets such as The New York Times and The Washington Post. Altogether, evidence has accumulated suggesting people tend to favor members of their own party on the one hand, and prefer to avoid members of the other party (and may discriminate against them) on the other.

But are people really choosing to socially avoid members of the other party for purely political reasons? While it’s possible that these decisions may be driven solely by partisan animus, another possibility is that that political affiliation is sometimes simply perceived to be correlated with relevant nonpolitical considerations that may factor into certain social decisions. For example, if someone reports that they do not want to have a member of the other party as a roommate, this need not be the result of differences in political opinions – this person might simply assume that they would not have much in common with an out-partisan in terms of relevant social, cultural, and lifestyle preferences. To illustrate the point, imagine a person who particularly hates country music, but is largely indifferent about politics. This person might report that they are unwilling to live with a potential roommate who is described as a Republican not because of their political views, but because Republicans (stereotypically) tend to enjoy country music.

To untangle partisan dislike from the use of partisanship as a cue for other relevant considerations, I use conjoint analysis to assess the impact of political and nonpolitical considerations on roommate selection. (Conjoint analysis is a technique commonly applied in marketing to assess which aspects of a product or service most heavily influence consumer preferences. Conjoints are similar to the vignette experiments familiar to many political scientists, but they allow researchers to vary many more treatment elements.) I asked college students to imagine they were taking part in a roommate selection process, and presented them with a series of side-by-side profiles describing hypothetical roommates which they were asked to choose between. These profiles included randomly varied information about each roommate’s cleanliness, preferred bedtime, personal values, social preferences, hobbies, musical taste, religion, sexual orientation, race, level of political interest, and partisan affiliation.

I find that partisanship strongly influences this social decision even in the presence of nonpolitical-but-politically-correlated individuating information. In other words, political affiliation isn’t just used as a stand-in for other characteristics (like music preferences) when it comes to social choices. Rather, people have powerful preferences based on partisan affiliation itself – strongly preferring to avoid members of the other party even when accounting for the cultural preferences their party membership may connote. In fact, when it comes to these kinds of decisions, partisanship may loom largest as a consideration – outweighing even more seemingly-germane factors for choosing a roommate such as their social preferences, preferred bedtime, or level of cleanliness. People preferred to avoid living with a member of the other party to a greater extent than they preferred not to live with someone who was described as “not at all clean and tidy”! All told, partisanship exerted a larger impact on roommate preferences than any other attribute.

Roommate-respondent attribute correspondence and preferences.Shafranek1.png


Contentious Federalism: Sheriffs, State Legislatures, and Political Violence in the American West

Zoe Nemerever, University of California, San Diego

Political violence surrounding federal land management policy provides a unique opportunity to advance understanding of both individuals’ decisions to engage in high-cost, unconventional political behavior and to explore the adverse consequences of federalist institutions.

Federal tensions over land and natural resources have been an integral part of politics in the western US. The federal government owns half of the land in the 11 western states (Arizona, California, Colorado, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming). Extensive federal land ownership limits the ability of states to determine how to manage their land for economic purposes, such as grazing and mining, in addition to recreational use and conservation.

Fig. 1 Map of federal land ownership in the western United StatesNemerever.png

As representatives of the federal government, employees of the Forest Service, National Park Service, Fish and Wildlife Service, and BLM routinely deal with politically-motivated threats, harassment, and physical violence from people upset about restrictions on the use of federal land. In a new Political Behavior article, I ask, what is the role of subnational governments in escalating political protest to political violence in modern United States?

I assert that political violence against federal bureaucrats is more likely when state legislation validates the views of those with complaints against federal land ownership and when county sheriffs signal that they hold anti-federal ideologies. First, the passage of land transfer legislation by state legislatures validates anti-federal political beliefs and can be perceived by citizens as a subtle endorsement for defying or harassing federal employees. Second, constitutionalist sheriffs elevate levels of political violence against federal employees by increasing the salience of anti-federal sentiments and lowering the costs of political violence. Constitutionalist sheriffs interpret the U.S. Constitution such that the federal and state government authorities are subordinate to county governments.

To substantiate these theories, I analyzed reports of violence towards the Bureau of Land Management. The reports include the date and location of 500 physical assaults, verbal harassment, and violent threats between 1995 and 2015.  Over two decades, 33% of western counties experienced civilian violence against Bureau of Land Management employees. I identified constitutionalist sheriffs from newspaper coverage and the websites of constitutionalist sheriff organizations. I also created an original dataset of land transfer legislation introduced in state legislatures.

In a multivariate analysis, I find that counties that elect constitutionalist sheriffs are more likely to have civilian violence against federal employees, and counties predisposed to political violence have higher rates of violence in the year following the passage of land transfer legislation.

The rate of violence was over twice as high in counties with a constitutionalist sheriff. Even after accounting for other factors, counties that elect constitutionalist sheriffs are 55% more likely to have political violence against federal employees. Additionally, the passage of land transfer legislation is associated with a 10% increase in the probability of political violence occurring in the following year.

An under-explored consequence of American federalism is one level of government encouraging civilian violence against another level of government. Incorporating federalism into the study of political violence illuminates how the actions of elected officials at state and county levels can promote violence against the federal government, a phenomenon previously unexamined by political scientists. Symbolic land transfer legislation and sheriff elections are often given short shrift by both media and academia, but these political activities have important consequences for the safety of public employees. Scholars should continue to examine other contexts of political violence to increase the discipline’s understanding of which subnational political institutions encourage violent political behavior, and how it can be avoided in the future.