PRL at NGA’s Disagree Better Convening

September 12, 2023 — PRL Director Sean Westwood presented on a panel in Manchester, NH, at the National Governors Association Disagree Better Convening. The event featured discussions with leading, bipartisan experts and focused on “Correcting Misperceptions and Highlighting Commonalities.” The NGA civility initiative, led by NGA Chair Governor Cox (R-UT) and Vice-Chair Governor Polis (D-CO), aligns with PRL’s research demonstrating that reducing partisan animosity and correcting misperceptions in America requires an elite-led approach. In fact, correcting misperceptions among citizens may first necessitate correcting a misperception among elites that negativity is an effective strategy.

Our work shows that public misperceptions are rampant. PRL data from our weekly America’s Political Pulse survey reveals that both Democrats and Republicans exaggerate the extent to which the out-party supports democratic norm violations and political violence. For example, Democrats and Republicans think nearly 60% of the opposing side would support a person charged with politically motivated assault on a member of their party (57.5% for Democrats and 60% for Republicans). The actual numbers are up to 30 times smaller–1.9% of Democrats and 3% of Republicans would actually support the person charged with the crime. The good news is that real support for norm violations and political violence aren’t as high as many perceive; the bad news is that misperceptions are helping to fuel partisan animosity in America.

PRL’s work is aimed at garnering a better understanding of why and how partisan animosity occurs. We currently capture weekly attitudes of Americans on affective polarization, norms, political violence, social trust, and more, and we are working on our second large-scale data project to classify elite rhetoric. Our research demonstrates that current levels of affective polarization are likely a response to elite rhetoric, but we know little about the dynamics of that relationship, including which politicians are most responsible for damaging language and why, and how rhetoric shifts in response to political events or during election cycles. 

The reality is that many elites and engaged citizens believe that “going negative” is the most effective strategy for winning an election. PRL recently conducted a review of the academic research on negative campaigning, which reveals that there is little evidence of a net benefit for politicians that attack fellow candidates, and it is certainly not worth the exorbitant costs associated with running attack ads. The research demonstrates that any small reduction in positive feelings for the subject of an attack are outweighed by reductions in positive feelings for the attacker. And even when small positive effects of negative campaigns are found, such as increasing awareness of an election occurring, these effects are case specific and short lived. 

For the majority of candidates, running negative campaign ads may appear strategic in the moment and not be motivated by any desire to harm society writ large, but the reality is that there is mounting evidence demonstrating that inflammatory elite rhetoric is indeed contributing to societal divides, as well as distrust in government and institutions.

The critical importance of good research and evidence is clear, and PRL’s future work studying elite communication and public attitudes will help concerned groups devise effective interventions. 

Not All Elections are Presidential: But They Might Be More Partisan Than Ever

Derek Holliday

Do presidential nominees dictate the fortunes of down-ballot candidates? While this has been a growing concern of many political commentators, I find evidence that elections for lower office are decided on partisan rather than presidential considerations.

Both Joe Biden and Donald Trump are viewed unfavorably by a majority of registered voters. Despite their unpopularity, both are likely to be their respective party’s nominee for the 2024 presidential election in a rematch of 2020 that few want to see again. This has Democrats and Republican political candidates for other offices worried; when voters go to the polls in 2024, will their electoral fortunes be tied to their party’s presidential albatross?

This concern is one shared by many political scientists, who have found evidence that the electoral fortunes of candidates for non-presidential offices have grown increasingly similar to those of their presidential co-partisans. As one political scientist put it, “all politics is presidential.” This has potentially serious consequences for political representation — if voters are making decisions for governor, for example, based on their views of the president, incumbent governors cannot be rewarded or punished for the actions they took over the past four years.

Are presidential candidates really to blame, or is some other force aligning presidential races with down-ballot results? To find out, I conducted a study comparing presidential influence to another major force in U.S. politics: partisanship. I hypothesized presidents are not primarily responsible for top-down influence of state and local elections. Instead, the growing importance of partisan identification more strongly structures behavior across all electoral contests.

Think of election results as numerical noise — a bunch of numbers with seemingly little pattern connecting them. To test my hypothesis, I horse-raced presidential voting against partisanship (which I measure as the general tendency for a place to vote for Democrats or Republicans, regardless of office), in a race to explain this noise. Once we account for either factor, how much noise is left? The more important the factor, the less noise there is remaining.

The figure above shows the relative predictive performance of partisanship versus presidential voting for all county-level results for U.S. Senate, Governor, State Secretary of State, and State Attorney General from 1972 to 2020 — positive numbers mean better partisan performance, negative mean better presidential. The red line shows that on average, county partisanship is a better predictor of results than presidential votes, suggesting politics have become more partisan than presidential.

Of course, these races tend to be of higher profile. Because of access to news coverage, larger advertising budgets, and greater name recognition, candidates for larger statewide office might be able to better distinguish themselves from presidential candidates. What about races for smaller offices? In the figure below, I compare the predictive power of presidential and partisan voting for every statewide contest between 2016 and 2020. Again, partisanship on average outperforms presidential voting in every office. In fact, the most obscure offices tend to have the largest partisan prediction advantage.

Candidates in down-ballot contests, then, are not just reflections of their co-partisan presidential candidates. They are, however, clearly identifiable as members of the same party. As partisanship becomes a greater dividing line in American life, candidates for lower office will have to work harder to distinguish themselves from other partisans, not just presidential candidates.

Holding Trump Accountable Has Not Threatened American Democracy

Derek Holliday, Yphtach Lelkes, and Sean J. Westwood


Did Donald Trump’s indictment increase support for democratic norm violations and political violence? Political commentators have indicated concern that pursuing criminal charges against the former president may inflame antipathy between partisans. Using survey data covering the periods before and after Trump’s indictment, we show no lasting, significant changes in the attitudes of partisans. Any differences in attitudes before and after the indictment quickly dissipated to pre-indictment levels within a few days. Our results have positive implications for the health of our democracy, suggesting politicians can be held accountable for criminal activity without widespread threat of retaliation from supporters.


Data


Our data come from a daily survey of Americans’ political attitudes. In total, we analyze the responses of 29,725 respondents from September 16, 2022 through June 29, 2023, which includes 1,763 responses gathered on or after June 14, the date of Trump’s indictment. All respondents in our data identify as either a Democrat or Republican and pass a simple attention check. All respondents are asked to rate their agreement with a series of five statements related to democratic norm violations and four statements related to political violence on a five-point scale from strongly agree to strongly disagree, which we recode individually as a binary measure of agree/don’t agree. We list the democratic norm items in full below:


1. Reduce outparty polling stations: Do you agree or disagree: (inparty) should reduce the number of polling stations in areas that typically support (outparty).2. More loyal to party than election rules and constitution: Do you agree or disagree with the following: When a (inparty) candidate questions the outcome of an election other (inparty) should be more loyal to the (inparty) party than to election rules and the constitution.3. Ignore outparty court decisions: Do you agree or disagree: (inparty) elected officials should sometimes consider ignoring court decisions when the judges who issued those decisions were appointed by (outparty) presidents. 4. President should circumvent congress: Do you agree or disagree: If a (inparty) president can’t get cooperation from (outparty) members of congress to pass new laws, the (inparty) president should circumvent Congress and issue executive orders on their own to accomplish their priorities.5. Censor partisan media: Do you agree or disagree with the following: The government should be able to censor media sources that spend more time attacking (inparty) than (outparty).
Additionally, respondents are asked several questions related to feelings of warmth toward members of the same party and members of the opposite party, with responses on a 0-100 scale. 


Method


We implement an interrupted time series design to measure changes in attitudes among partisans pre- and post-indictment. Specifically, we assume independent trends in support for democratic norm violations, political violence, and affective polarization before and after Trump’s indictment. Over-time trends, however, can be linear or non-linear. In our preferred specification (shown in the graphical results), we fitted smoothed (loess) curves over the pre- and post-indictment periods. This allows for the trend to more closely follow short-term increases and decreases in our variables of interest. In the appendix, we fit linear models to the data, which largely cohere with the models we present below. All analyses are weighted to reflect the national population.

Results


First, we analyze changes in support for political violence between May 1 and July 26, 2023, by party. Specifically, we estimate the proportion of respondents supportive of assaulting an out-party member. We find there is no meaningful change in support for violence after Trump’s indictment. Such support remains incredibly rare (around 3 percent) among both Democrats and Republicans. 

Support for political violence, of course, represents one of the more extreme attitudinal positions respondents can take. How did support for democratic norms change after Trump’s indictment? We measure support for democratic norm violations as an index, summing over the five norms in our survey and reporting the average number of norm violations supported by partisans. Here, we find there is a very small increase in support for norm violations among Republicans immediately after the indictment of around one seventh of a single additional norm violation, although Republicans still do not become more supportive than Democrats.* Additionally, this small increase in support vanishes extremely quickly, returning to baseline levels just a week after the indictment. This suggests attitudes toward democracy are largely robust to Trump’s legal battles.

* We caution, however, against overstating the significance of apparent differences in loess curves that emerge from daily variations in attitudes. While our sample is large in aggregate, there is randomness in the representativeness of respondents completing the survey on any given day. While we omit confidence intervals in our graphics to more clearly show the trend lines themselves, we note here many of the apparent effects in the days immediately following Trump’s indictment are highly uncertain.

We also aim to capture more generalized changes in partisan feelings toward both in-party and out-party members. We may expect feelings toward rival partisans to grow colder after Trump’s indictment, which could further chill partisan relations in the country. We test this possibility by tracking the level of affective polarization by party pre- and post-indictment. Affective polarization is defined as the difference in feelings of in- and out-party warmth on 0-100 feeling thermometers.


Similar to trends in support for democratic norm violations, we see a small uptick in affective polarization among Republicans after Trump’s indictment, but this small increase rapidly falls back to pre-indictment levels. At present, partisans are again at parity with regard to affective polarization.

Surprisingly, however, this temporary increase in affective polarization was not driven by out-party antipathy. As we show below, it was in-party affect that increased (among both parties) after the indictment. Among Republicans, this affective bonus toward fellow Republicans dissipated quickly back to pre-indictment levels, which we interpret as a sort of “rally around the flag” effect. Democrats, on the other hand, showed a slightly more temporally robust increase in warmth toward co-partisans. This may be interpreted as an expression of self-congratulation for finally bringing criminal charges against Trump.

Conclusion


Trump’s indictment did not seem to change Americans’ attitudes about political violence, democracy, or members of the opposing party. If any changes did occur, they were both minor and fleeting. We take this evidence as hopeful for the prospects of holding public officials, even of high importance to partisans, accountable for their actions. While we cannot rule out small, radical groups retaliating for legal prosecution of Donald Trump, it is certainly not the case that broader American attitudes changed in the wake of his indictment.


Appendix


For all models in the main text, we make minimal parametric assumptions across all attitudes by fitting loess curves to the pre- and post-indictment periods. This makes the curve sensitive to local minima and maxima in the data, as the slope of the curve is weighted to be responsive to local deviations. 


In addition to the loess models shown above, we fit a series of corresponding linear specifications and report our results below. For each dependent variable, we estimate two specifications. First, we regress our dependent variable on a binary indicator for pre- or post indictment, a dummy for Republican party identification, and an interaction between the two. The quantity of substantive interest is the coefficient on the interaction term, indicating whether average Republican attitudes changed post-indictment.


For our second specification, we split the post-period into the first and second weeks post-indictment in order to determine shorter-term trends in the post-indictment period. The quantity of interest is again the coefficient for the interaction term. If changes in Republican attitudes are greatest in the week immediately following Trump’s indictment, we’d expect the coefficient on the interaction between week 1 and Republican to be significant and the interaction for week 2 to be insignificant.
 

Table: Support for Assault

OLS Pre-PostOLS Weeks
(Intercept)0.034***0.034***
(0.001)(0.001)
Post Indictment0.002
(0.006)
Republican-0.010***-0.010***
(0.002)(0.002)
Post x Republican0.004
(0.008)
Week 1 Post-Indictment0.005
(0.008)
Week 2 Post-Indictment-0.001
(0.007)
Week 1 x Republican0.004
(0.012)
Week 2 x Republican0.005
(0.011)
:———————–————-:———-:
Num.Obs.2972529725
R20.0010.001
Log.Lik.8949.5958949.794


 Table: Support for Democratic Norm Violations

OLS Pre-PostOLS Weeks
(Intercept)1.016***1.016***
(0.011)(0.011)
Post Indictment-0.093*
(0.041)
Republican-0.154***-0.154***
(0.016)(0.016)
Post x Republican0.165**
(0.062)
Week 1 Post-Indictment-0.119*
(0.060)
Week 2 Post-Indictment-0.072
(0.055)
Week 1 x Republican0.195*
(0.090)
Week 2 x Republican0.141
(0.083)
:———————–————-:———–:
Num.Obs.2591525915
R20.0040.004
Log.Lik.-43884.794-43884.612

Table: Affective Polarization and In-Party Warmth

Affpol Pre-PostAffpol WeeksIn-Party Pre-PostIn-Party Weeks
(Intercept)54.093***54.093***75.692***75.692***
(0.259)(0.259)(0.169)(0.169)
Post Indictment-1.620-0.267
(1.047)(0.682)
Republican-1.680***-1.680***-1.781***-1.781***
(0.385)(0.385)(0.250)(0.250)
Post x Republican0.6620.024
(1.584)(1.031)
Week 1 Post-Indictment-0.1881.189
(1.524)(0.993)
Week 2 Post-Indictment-2.819*-1.486
(1.399)(0.911)
Week 1 x Republican-1.528-1.574
(2.298)(1.493)
Week 2 x Republican2.5151.367
(2.126)(1.385)
:———————–—————-:————-:——————:—————:
Num.Obs.29574295742968029680
R20.0010.0010.0020.002
Log.Lik.-145893.953-145892.932-133700.579-133698.529

Note: * p < 0.05, ** p < 0.01, *** p < 0.001
Note: Estimated with survey weights

Introducing the Library of Partisan Animosity

The Polarization Research Lab announces the launch of the Library of Partisan Animosity, an information and learning hub for the most important academic work on partisan animosity–the study of what drives people to like the political party they identify with and dislike the other party, and the consequences of that gap.  

A breakdown in the dissemination of political science research to practitioners, media, and citizens often occurs due to a lack of a centralized hub, inaccessible academic language, and journal articles living behind expensive paywalls. The Library of Partisan Animosity is an unprecedented effort to bridge this gap by creating a free resource that is easy to navigate and clearly translates the key research and findings. With the most cited and innovative work on partisan animosity residing in one location, users can for the first time follow the trajectory of the scientific work on these topics and, importantly, discern where the gaps in our understanding of partisan discontent still lie. 

The Library of Partisan Animosity will organize and summarize 100+ papers on political polarization this year, allowing anyone with interest to freely access the learning from this research. We expect the library to be particularly useful for those without regular institutional access to academic work, including K-12 educators, the media, and practitioners. The Library can help users explore questions such as: What drives political polarization? How does one’s identity affect partisan views? Does this animosity translate to support for undemocratic actions?  

The summaries are written by academics and advanced graduate students with deep knowledge of American politics and research methodology. While academics have studied partisanship for decades, the literature on partisan animosity over the past 10 years has expanded, in large part due to advances in experimental design and survey methodology. The Library of Partisan Animosity summaries strive to help readers understand the strengths and limitations of the experiments and methods used in the papers. 

Each Library summary includes: 

  • Explanations of the contribution that the paper makes in our understanding of partisan animosity
  • The key findings of the study
  • Strengths and weaknesses of the methods and data used 
  • Good research “best practices” checks 
  • The ability to filter by year and topic (tags) 
  • Definitions of key terms

The first five summaries of seminal work in the field of partisan animosity are now available, and we will continue to add to the Library over the coming months. 

Tracking America’s Political Pulse

The Polarization Research Lab has launched a new visual dashboard for understanding partisan animosity in America. For the first time, anyone can view accurate and up-to-date polling on the state of polarization and democracy in America just by visiting the free and public PRL website. 

Using data from the Lab’s weekly tracking survey of 42,000 responses (and counting) since October 2022, the dashboard provides gauges on:

  • Levels of polarization over time by party and state 
  • Attitudes toward trust (in institutions and elected leaders)
  • Support for democratic values
  • Support for violations of democratic norms
  • Support for political violence 

Viewers can easily track trends in these measures across states in map form, by total count, or by party in easy-to-read visualizations. 

With our publicly available data and new dashboard, PRL hopes to empower citizens–voters, students, practitioners, members of the media, and elected officials–with the information needed to understand the status of American democracy from the national to local level. 

The dashboard includes scores and rankings of states for levels of polarization and support for democratic norm violations. In the future, as we collect more data, we will provide this detail down to the county level, making the dashboard an invaluable resource for community groups and local leaders seeking to understand their constituents. 

Existing surveys that reach a large number of Americans only do so at incremental points in time, and small polls often fail to interview sufficient numbers of respondents to generalize the findings. As a result, citizens, the media, and our elected officials respond to this piecemeal data, which does not accurately reflect Americans’ attitudes, leaving us with messaging and information that at best is inaccurate and at worst aggravates polarization. 

America’s Political Pulse fills this critical gap in the data available on partisan animosity by collecting daily responses over 3+ years. This level of granularity allows PRL for the first time to examine the effects on polarization and democracy of political moments in real time, such as the midterm elections, court rulings, and the indictment of President Trump. 

Having consistent, accurate data is essential to understanding the strengths of American democracy as well as the areas of weakness, and is the only viable first step in identifying the causes of partisan polarization and developing possible solutions.  

Introducing the Pledge of Civility: An Audited Pledge from Future Elected Officials

The Polarization Research Lab has launched the Pledge of Civility, a new initiative aimed at lowering political hostility in America with an audited pledge from our future elected officials and staffers. The need for such an effort is clear: researchers at the Lab have shown that partisan animosity and affective polarization–the difference between feelings toward the in-party versus the out-party– have increased steadily in the last 30 years, seeped in areas of life outside of politics, and show little signs of change since the Lab began collecting data (see America’s Political Pulse). 

One of the goals of the Polarization Research Lab is to bring together academic research and real-time interventions to help ameliorate partisan animosity in America. The Pledge of Civility is the first such effort that merges an evidence-based, elite-level intervention to counter partisan animosity with an auditing mechanism. 

The simple pledge focuses on positivity and avoiding language that polarizes citizens: 

“I pledge to choose civility when campaigning,

  • I will focus on issues and civil debate.
  • I will not engage in name calling or personal attacks on the candidate or their family.  
  • I will avoid attack ads whenever possible.”

Our target audience for the Pledge is anyone running or planning to run for office at any level—local, state, national—in the next decade.  We hope to reach candidates early in their political careers to invite credible commitments to civility and positivity in campaigns. The Polarization Research Lab and the Pledge of Civility are nonpartisan; we are an academic research Lab interested in understanding partisan animosity and how and why interventions may affect political polarization. 

We believe that two key aspects of our pledge set us apart. First, we are targeting the future political elites of America at the local, state, and national level, rather than citizens. Our focus on elites is founded on our academic research demonstrating elite rhetoric shapes citizen attitudes. We need to capture candidates early, typically before they have run for office, for candidates to credibly commit to civility.  

Targeting those who hold or will hold political power–known as “political elites”–for the Pledge of Civility stems from years of academic work in political science from PIs at the Lab showing that elite hostility and negative campaigning increase the likelihood that citizens also become more negative. Most alarmingly, political conflict among elected officials not only causes average citizens to dislike the other side, but it skews behavior in entirely apolitical realms (dating, hiring, medical care, etc.). By running a civil campaign, candidates signal to voters that the out-party is not deserving of hate and candidates can focus on political issues.

Our second key contribution is that we are providing what we believe is the first audited pledge of this kind. Most pledges have no way of showing effectiveness or credible commitment by signatories because they do not track and measure elite language. The Polarization Research Lab uses machine learning techniques to collect and code elite rhetoric. We currently capture elite language through tweets, floor speeches, press releases, and newsletters and this language is then coded for the presence of polarizing and depolarizing language. All of our data is publicly available on our website, and we will provide data about the use of polarizing rhetoric for both elites who sign the pledge and those who do not. We expect this data to be very useful for the media and any organization or individual hoping to track and hold accountable elites for their commitment to civility and depolarization. 

We invite candidates over the age of 18 to sign the Pledge of Civility. To disseminate information about the Pledge, we hope to collaborate with public policy and law schools, where many of the next generation of political elites are currently studying. In addition, we plan to build relationships with the vast network of nonprofit and professional organizations that find and train future elected officials, particularly at the local and state level. We see this as a unique opportunity to bridge the academic-community divide and truly test and measure an intervention founded in science and put into practice everyday by educators, practitioners, and political activists. 


To learn more about partnership and how to share information about the Pledge of Civility, please contact Emma Cutrufello at emma.r.cutrufello@dartmouth.edu. References for the research cited above are available on the Pledge of Civility webpage.

PRL First Annual Meeting Convenes Top Scholars on Partisan Animosity

The Polarization Research Lab hosted its First Annual Meeting in collaboration with the Hoover Institution on March 2-3, 2023. Fifty-five attendees enjoyed an instructive two days of research and learning about partisan animosity at the Hoover Institution’s beautiful space in Palo Alto. 

The meeting brought together academic experts and graduate students across disciplines studying partisan animosity in the US and abroad, practitioners working with communities around the country to temper political polarization, and funders. The goal was to convene these groups in-person to bridge the academic to practitioner divide and share advances in the study of partisan animosity among the top researchers in the field. Few opportunities exist for young researchers and practitioners to attend a conference with senior faculty from political science, psychology, sociology, and economics, and PRL designed the conference to maximize opportunities for networking and mentoring.  

On March 2, Tad and Dianne Taube Director of the Hoover Institution Condoleezza Rice, welcomed the attendees and introduced keynote speaker Utah Governor Spencer J. Cox. Dr. Rice and Governor Cox both spoke to the salience of a gathering on reducing partisan animosity and the critical importance of social science research in uncovering causes and effective treatments to reduce toxic polarization. Speaking about his campaign, Governor Cox highlighted his and his opponent’s efforts to campaign civilly and positively while maintaining intellectual debate about policy.    

The following two days included seven academic talks with discussant feedback and five panel presentations, all of which induced highly engaged audience questions. The meeting also introduced a venue for researchers to present their works in progress in a small group workshop format. For early-mid stage projects, this was a unique opportunity to receive cross-disciplinary feedback from the foremost scholars of partisan animosity and will help these projects get published in top journals. Fifteen poster presentations by graduate students and postdocs during a cocktail reception before a catered dinner rounded out a highly successful first day of the meeting. 

By the end of the second day of presentations, it was clear that the meeting succeeded in its goal of advancing our understanding of partisan animosity and laying a path for future work. A central theme of the research presented was ensuring that we understand the causes of partisan animosity, and polarization before we devise treatments, so that we are sure we are correcting the actual problem and not just a symptom of the problem. Presentations covered topics including what drives partisan animosity and how do we define it–it’s both identity and policy issues–and the extent and measurement of democratic backsliding in the United States–which may not be as severe as some discourse suggests. 

Another highlight was presentations and posters that included initial results from experiments designed to correct misperceptions and depolarize study participants. While academic conferences often dwell on the theoretical, the PRL meeting prioritized discussions that advance our understanding of how to effectively treat and scale polarization interventions. Again, clear themes emerged, and robust dialogue centered on the specific conditions under which misperceptions, particularly toward the out-party, can be corrected as well as how journalists can help counter misinformation in the way they frame story headlines. 

Researchers also presented initial findings from studies that used data from PRL’s America’s Political Pulse survey, the Lab’s signature data collection effort to track partisan animosity through weekly surveys. As part of its core focus on providing public goods and advancing our understanding of political polarization, the Lab offers survey time free of charge to researchers through an RFP and has already fielded 17 surveys with another 10 accepted this winter. The high-quality research presented at the meeting showcased the importance of well-designed and replicable experiments in understanding and treating partisan animosity.   

In his opening, PRL Faculty Director Sean J. Westwood spoke about new projects coming from the Lab in 2023, including the Library of Partisan Animosity and the Pledge of Civility. The Library is a core public good provided by the Lab and will provide summaries of the seminal works on partisan animosity as well as information about experimental design and methodology. Covering more than 100 articles it will be an indispensable resource for students, educators, practitioners, and the media. The Pledge of Civility is an elite-level intervention designed by the Lab aimed at lowering political hostility in America with a simple audited pledge–”I pledge to choose civility when campaigning”– from future elected officials and staffers. 

The Polarization Research Lab is funded by the Koch Foundation, New Pluralists, and the Knight Foundation. The Lab was founded in 2022 by Sean J. Westwood, Associate Professor  at Dartmouth College, Yphtach Lelkes, Associate Professor at the University of Pennsylvania, and Shanto Iyengar, Professor at Stanford. Questions or media inquiries can be directed to Sean J. Westwood at sean.j.westwood@dartmouth.edu.