In the spring of 2020, I spoke to Danielle, a white woman living in Iverson, Wisconsin,Footnote 1 about how the federal government was handling the COVID-19 pandemic. She explained that she was struggling to form an opinion about the subject because she wasn’t sure what the government actually had and had not done. As she told me, “Honestly, it’s all hearsay. I don’t really know how the government is handling it… . I tried to do my own research … other than, oh, what does CNN say, oh, what does Fox News say? I try not to take their word for what’s actually happening.” Contrast Danielle’s confusion to the certainty that Rose—another Iverson resident—expresses around the same time: “I think (Trump) was more worried about the economic fallout since it’s an election year for him. I think it was downplayed.”
Rose and Danielle are both white women living in the same community who have always voted for Democratic presidential candidates, though Danielle prefers the label Independent. What differentiates the process by which they arrived at opinions about COVID-19 policies is less their partisanship than their trust in media: Rose believes what she sees on CNN or ABC, whereas Danielle feels the need to “do her own research.” Republicans I spoke to around the same time reached different conclusions but by a process that was similar to Rose’s—self-selecting into a certain media environment that reinforced their opinions.
Political scientists have documented rapid issue polarization among both elites and the mass public on COVID-19 mitigation policies and public health behaviors (Allcott et al. Reference Allcott, Boxell, Conway, Gentzkow, Thaler and Yang2020; Kushner Gadarian, Goodman, and Pepinsky Reference Gadarian, Shana and Pepinsky2020; Motta, Stecula, and Farhart Reference Motta, Stecula and Farhart2020). But these partisan differences cannot explain someone like Danielle, who simply lacked confidence in her opinions because she lacked trust in the media. This raises the question, how does trust in media shape the process of opinion formation?
This paper offers unique insight into this question, drawing on four rounds of semistructured, in-depth interviews with 86 voters living in three states across the Midwest. The interviews were conducted over the course of 18 months: the first during the summer and fall of 2019; the second in February and March 2020; the third in April and May 2020; and the fourth between September and November 2020. Thus, the data capture participants’ trust in media, information gathering practices, and political opinions before the pandemic, as well as during periods of uncertainty and partisan division during the pandemic.
Among my participants, I found distinctive trajectories of opinion formation on COVID-19 issues as they confronted the complex information environment armed with varying levels of trust in media and political knowledge. The first trajectory mirrors the survey results described above: rapid polarization about all matters related to COVID-19. But this occurred only among participants who had a trusted—usually partisan—news source and sufficient knowledge of partisan interpretations about the pandemic to follow them (Slothuus and de Vreese Reference Slothuus and de Vreese2010).
Among participants without a trusted media source, I identified two trajectories of opinion formation. Those who believed that all media reported some biased version of truth but had sufficient political knowledge to triangulate among those biases were often able to reach some conclusions about the government’s response to the pandemic. But among those who lacked all trust in media, the parties’ conflicting interpretations of reality made for difficult terrain to navigate. Without a trusted media outlet to guide them to the “right” answers, these participants often ended up opinion-less when it came to COVID issues. By the eve of the 2020 presidential election, the aggregate result was a striking combination of polarization and disengagement among this sample of white Midwesterners.
By taking an inductive approach to understanding how participants both gathered and interpreted information about the pandemic, this study offers two contributions to research on the relationship between media consumption and issue polarization. First, along with others (Levendusky Reference Levendusky2013), I suggest that the media plays a role in polarization primarily because of extremism at the tails of public opinion rather than division at the center. But in contrast to other accounts, I argue that some portion of those at the center might actually be relatively opinionless rather than “centrist.”
In so doing, I build on experimental studies that have sought to imitate real-world settings in which participants are able to select information at different points (Druckman, Fein, and Leeper Reference Druckman, Fein and Leeper2012). Although my findings cannot determine whether population-level behaviors are best characterized by partisan news-watching (Jamieson and Cappella Reference Jamieson and Cappella2010; Stroud Reference Stroud2008), the consumption of multiple news sources from competing perspectives (Gentzkow and Shapiro Reference Gentzkow and Shapiro2011), or political disengagement (Prior Reference Prior2007), they suggest that a crucial piece of variation in these behaviors and their consequences for opinion formation is about trust in political information.
And although the kind of trust that mattered most among my participants was trust in the media, distrust in other information sources—whether the government, “the experts,” or politicians themselves—may produce the same lack of opinion formation. Therefore, increasing attacks on all kinds of knowledge-producing institutions (Rosenblum and Muirhead Reference Rosenblum and Muirhead2020) suggest that trust may be an increasingly important factor in the way that people gather and interpret political information—particularly about novel issues.
Political Knowledge and Opinion Formation about Novel Issues
Existing accounts of opinion formation suggest that even small biases in the information environment when a novel issue is presented—either through framing effects or people’s selective media exposure—can cause rapid divergence in public opinion (Druckman and Bolsen Reference Druckman and Bolsen2011). As soon as people catch hold of a partisan frame, politically motivated reasoning takes hold (Kunda Reference Kunda1990; Lodge and Taber Reference Lodge and Taber2013).
Politically motivated reasoning contains two distinct but related components: people process information differently, crediting information that corresponds to their predispositions—namely, partisanship—and discrediting information that does not; moreover, when given the opportunity, they also gather information differently, selecting information that confirms their existing attitudes (Taber and Lodge Reference Taber and Lodge2006). And because U.S. media offer citizens so much choice, they can choose to consume only partisan news sources (Jamieson and Cappella Reference Jamieson and Cappella2010; Prior Reference Prior2007).
Furthermore, as information gathering and processing play out over time, existing accounts suggest that they interact to produce increasing opinion polarization: people dig in their heels in defense of early opinions as they repeatedly reselect information favorable to those opinions (Druckman, Fein, and Leeper Reference Druckman, Fein and Leeper2012).
But to engage in this polarizing process of information gathering and processing, citizens must first have sufficient political knowledge to recognize different positions as partisan (Layman and Carsey Reference Layman and Carsey2002; Slothuus and de Vreese Reference Slothuus and de Vreese2010). Only then can they seek out material to contest facts (Bartels Reference Bartels2002; Flynn, Nyhan, and Reifler Reference Flynn, Nyhan and Reifler2017) or reinterpret factual accounts to support their party’s position (Gaines et al. Reference Gaines, Kuklinski, Quirk, Peyton and Verkuilen2007). Political knowledge is likely to be particularly important when it comes to new issues, as only the politically savvy will be able to track elite polarization (Layman and Carsey Reference Layman and Carsey2002).
Even this insight, however, only accounts for one facet of political opinion formation. Articulating a political opinion requires two components: a degree of certainty in some opinion and a way to relate that opinion to the political landscape.Footnote 2 Political knowledge takes care of the latter, but as I will show trust in political information is essential for the former, particularly when it comes to novel issues.
Trust and Opinion Formation
Political trust is endogenous to institutional contexts: citizens assess how well institutions perform relative to their expectations, develop trust (or not) in those institutions and decide whether and how to engage with them (Hetherington Reference Hetherington2005; Nannestad Reference Nannestad2008; Paxton Reference Paxton2002). Thus, to understand how (dis)trust in sources of political information might shape opinion formation, we must begin with how the U.S. information environment produces or fails to produce trust. In the case of COVID-19, as for many issues, political information was conveyed to the American public through a number of overlapping channels: government agencies, party elites, public health experts, and the media.
Moreover, these information channels, are suffering from a shared crisis of legitimacy. As populist appeals target out-of-touch government bureaucrats, party elites, and all numbers of experts (Bonikowski and Gidron Reference Bonikowski and Gidron2016; Brubaker Reference Brubaker2021; Cramer Reference Cramer2016) and the “new conspiracism” seeks to discredit a wide swath of knowledge-producing institutions (Rosenblum and Muirhead Reference Rosenblum and Muirhead2020), there are few sources of political information whose trustworthiness has not been called into question in recent years.
And these changes seem to have had consequences for citizens’ trust. According to Gallup, only 32% of Americans claimed to have a “great deal” or “fair amount” of trust in the media in 2016 (Swift Reference Swift2016), and in 2019 Pew reported that 69% of Americans believed the government withholds information unnecessarily (Perrin and Rainie Reference Perrin and Rainie2019). Amidst the COVID-19 crisis, just above half of Americans had a great deal of trust in the health recommendations of the CDC (52%) while only 37% had similar levels of trust in the FDA and the NIH (Robert Wood Johnson Foundation/Harvard School of Public Health 2021).
But these declines are not even across the population; Americans tend to evaluate a source’s credibility based on the extent to which it agrees with their political predispositions (Zaller Reference Zaller1992), and conservatives in particular report a growing distrust in the “liberal bias” of “mainstream media” (Lee Reference Lee2010). Trust in expertise has also become an increasingly partisan phenomenon. Between 1974 and 2010, political conservatives went from the most to least trusting in science, relative to both moderates and liberals (Gauchat Reference Gauchat2012). There is also evidence that this trend is linked to support for populist policies, driven by an underlying predisposition toward anti-intellectualism (Merkley Reference Merkley2020; Motta Reference Motta2018). And although this kind of anti-intellectualism, populism, and even conspiracism are not necessarily partisan, they often find fruitful ground to flourish amidst Republican Party elites who either do nothing to combat these trends or wield them for their own political ends (Rosenblum and Muirhead Reference Rosenblum and Muirhead2020).
In summary, this evidence indicates that trust in sources of political information may itself become a partisan phenomenon. But we also know that some Americans—likely those already polarized—engage in selective exposure, only consuming partisan news media that polarizes them further (Jamieson and Cappella Reference Jamieson and Cappella2010; Levendusky Reference Levendusky2013; Stroud Reference Stroud2008). This suggests that although some people may genuinely distrust all forms of political information, the most polarized trust selectively: they trust partisan news media.
Trajectories of Opinion Formation
Politically motivated reasoning aptly describes the process of political opinion formation for this polarized group. They know which information sources are partisan and trust in-party sources when seeking out information on novel issues; they use that information to form an opinion that supports their party; and eventually, they may become “even more extreme” (Levendusky Reference Levendusky2013, 612). Among my sample of Midwesterners, I refer to this group as “trusted sourcers.” They are defined, in an era of increasing distrust in political information by the fact that their distrust is selective. Although they may claim to distrust all political information, in reality they trust in something, often partisan news media. This selectivity, rather than political ideology, is what distinguishes them from their peers in the process of opinion formation.
But what about those whose distrust is more universal? This could include people who have little trust in any media, the government, or experts—a confluence that the literature suggests is relatively likely as these institutions have come under similar lines of attack. For people with universal or near-universal distrust, the process of political opinion formation breaks down from the start. Reaching any opinion about a new issue may be challenging if one distrusts all sources of information. But among my interviewees, there is important variation within this group. Those that I call “triangulators” have more political knowledge and a minimum amount of trust in media, whereas those to whom I refer as “nonbelievers” have lower levels of knowledge and trust.
Triangulators have important stores of political knowledge, often a map of which media outlets they view as partisan and some information about what issue positions the parties support. This partisan map, along with the belief that discounting for a known bias in media reporting will reveal an underlying truth, allows triangulators to form an opinion about novel issues and then connect that opinion to the political system. Nonbelievers are in the most challenging position. A lack of political knowledge and universal distrust coincide such that this group has no sense of how to weigh competing claims about reality. They struggle the most to even form an opinion about a novel issue, let alone one that can be connected to the political system.
In the remainder of the paper, I show how trust in media and political knowledge shape the information gathering and interpretation processes among my participants throughout the pandemic. As I detail below, what mattered less in shaping these divergent trajectories of opinion formation was trust in government and experts. Within this sample of Midwesterners, people often trusted that the government was making a good-faith effort to collect data on COVID-19 deaths and cases such that many partisans agreed on the facts but disagreed on their interpretation (Gaines et al. Reference Gaines, Kuklinski, Quirk, Peyton and Verkuilen2007).
Data and Methods
Data Collection
This study draws on 293 in-depth interviews with 86 residents of sparsely populated counties in the Midwest that were conducted between June 2019 and November 2020. The data are part of an ongoing study of these communities, which I refocused to incorporate participants’ assessments of the pandemic response.
The larger study examines the relationship between place and political behavior, asking, what is it about community life that leads similar people to vote differently? Therefore, the counties were selected based on their demographic similarities and partisan differences. All three are composed of predominately white residents employed in education, manufacturing, transportation, and service sector occupations. Despite these similarities, they have voted differently in presidential elections since the 1960s: Iverson, Wisconsin, is Democratic; Meriville, Indiana, is Republican; and Williston, Minnesota, could go either way in any election.
The study focuses on white, industrial communities in the Midwest because these places are sites of interest for scholars studying populism, conservatism, and political subjectivities among the white, working classes (Cramer Reference Cramer2016; Gest Reference Gest2016), and the larger study was interested in examining local variation in white, working-class political alignments.
After identifying a set of potential cases that matched on demographic axes but diverged politically, I narrowed the sample to focus on counties centered on small towns with between 16,000 and 28,000 residents. As urban centers have become monolithically Democratic and rural areas monolithically Republican (Nall Reference Nall2015; Rodden Reference Rodden2019), it is increasingly rare to find political differences among similar kinds of places. Small towns offer this kind of heterogeneity.
During the first phase of data collection in summer 2019, I spent four to six weeks in each community, living with residents, observing political and civic activities, and recruiting participants.Footnote 3 To do so, I used a variety of methods. I met people through community organizations, Facebook groups, and flyers, as well as in coffee shops, the YMCA, the public library, and at community events. I also recruited through snowball sampling, targeting people who fulfilled certain characteristics. I sampled purposively (Small Reference Small2009) with two objectives. First, I sought sufficient variation in gender, age, occupation, education, partisan affiliation, and political knowledge; second, I sought to recruit similar samples in each community for the sake of comparison.
Table 1 shows summary statistics for the sample’s demographic and political characteristics, and Table A1 of the Appendix provides individuals’ details and pseudonyms.
Table 1. Sample Political and Demographic Characteristics
a Based on the author’s qualitative coding, respondents were ranked on a scale from 1 to 4, where 1 = little sense of what differentiates the major parties, does not watch the news, may or may not vote; 2 = some sense of what differentiates the major parties, may watch the news, votes with regularity; 3 = knowledgeable about partisan differences, watches the news regularly; 4 = avidly attuned to politics.
There are two features of the sample that are relevant for evaluating the analyses presented here. First, I focused on achieving variation in political information and partisanship within each research site. Table 1 summarizes participants’ average political information, based on my qualitative coding. In many instances, the residents who responded to direct solicitations on Facebook or posters that were hung around town did so because they were politically engaged, so I relied on snowball sampling to reach those residents who otherwise might not like to “talk politics.” I often did so by explicitly asking participants to direct me toward their “least political” acquaintances. Second, the sample is composed of only white participants. This was intentional, as the larger study originally focused on how the communities shaped the development of white identity and political subjectivity. Therefore, all three counties are overwhelmingly white.Footnote 4 I discuss the limitations of this sampling strategy below.
I conducted the first set of interviews in person during the summer and fall of 2019. They lasted 1.2 hours on average. I completed the second round of interviews in Williston in late February and early March 2020, at which point COVID-19 made in-person data collection impossible. I switched to phone and Skype interviews for the remainder of the data collection and completed the second interviews in Iverson and Meriville during March and early April, the third in late April and May, and the fourth between September and November 2020. All participants received and signed a written consent form and notification of voluntary participation during the first in-person interview. In each following interview, I followed a verbal consent procedure. The retention rate between the first and fourth interviews was 80%.Footnote 5
Data Analysis
Two aspects of these data make them ideally suited to answering questions regarding public opinion formation about a novel issue. The first is that the timing of fieldwork offers insight into participants’ trust in media and political opinions prior to the pandemic as well as at three points during the pandemic—including right before the 2020 election. Although attitudes continued to evolve after the election, understanding how people arrived at the opinions that informed their vote choice is important in and of itself.
Second, in-depth, semistructured interviews are well suited for drawing comparisons between “contexts, situations, and kinds of people” (Lamont and Swidler Reference Lamont and Swidler2014). This analysis centers on comparison among people with varying degrees of political knowledge, partisan attachments, and trust in media. Therefore, interviews helped shed some light on contradictions between participants’ expressed distrust in the media and actual media-viewing practices.Footnote 6
Qualitative methods also allow for a combination of inductive reasoning and deductive hypothesis testing (Timmermans and Tavory Reference Timmermans and Tavory2012), which informed my approach to both data collection and analysis. After COVID halted in-person interviews in spring 2020, I incorporated new questions about the pandemic and public health policy into my interview guides for the second- and third-round interviews. I then transcribed the interviews using a combination of voice-to-text software and human editing done by myself and research assistants. I analyzed the transcripts in MAXQDA using an open coding approach, searching for unexpected and emergent patterns in the data (Saldaña Reference Saldaña2009). I completed all coding myself. As I did so, I found that a number of people were so distrustful of political information that they rarely arrived at an opinion about COVID-19 without second guessing themselves.
To assess this further, I developed finer-grained questions that adjudicated between trust in different sources of political information—media, government, and experts. Specifically, I asked whether participants had found any particularly trustworthy source of information about the pandemic, how trustworthy they found the media in reporting about the pandemic, how trustworthy they found the federal government in reporting data about the pandemic, and how they felt about Drs. Fauci and Birx. When participants did express distrust in media, experts, or the government, I probed those expressions in detail.Footnote 7
Based on analyses of these data, I inductively developed a typology of participants’ distrust and information gathering strategies to understand how these connected to opinion formation. I then returned to my first set of interviews, assessing the extent to which that typology fit participants the year before. Was this a phenomenon brought on by the pandemic? The answer was no. Rather, the pandemic made clearer the differences among people with and without trusted sources. This is because participants with universal distrust still form some political opinions, often through conversations with family and friends or experiences within their communities, but these were harder to come by with such a new and complex issue. As in other settings, the exceptional circumstances of the pandemic revealed social processes that were otherwise difficult to uncover (Klinenberg Reference Klinenberg1999).
Positionality
My approach to each field site and the data I collected in conversations with participants are shaped by my positionality, which is marked by the intersection of my multiple identities as a young, middle-class, white woman in graduate school at an elite, urban university (Collins Reference Collins2015). The salience of these identities also shifted in interaction with different participants, often when they asked me questions about myself (Reyes Reference Reyes2020). Older participants wondered about my religion (I attended a Catholic school) and my relationship status (I was in a committed heterosexual relationship), and many people asked where I grew up (a small town on the East Coast). Often, my membership in the Millennial generation and the fact that I was still in school in my mid- to late-20s obscured my class position, as middle-class and elderly participants regularly fretted over my student loans and job prospects, whereas other Millennials commiserated over monthly loan payments with me. Thus, these identities occasionally combined to afford me moments of “insider” status in which participants aligned themselves with me based on shared racial, generational, or religious identities and small-town experiences (Reyes Reference Reyes2020).
But in many ways, I was a clear outsider. I was still an academic and an “urbanite” studying small towns and rural communities. Often, these differences proved analytically useful, as residents explained facets of small-town life and rural–urban differences that they assumed I would not understand (see Cramer Reference Cramer2012). But these differences also indicate the power relationships between participants and myself, as I am the person who controls the writing of others’ stories (Cobb and Hoang Reference Cobb and Hoang2015; Small Reference Small2015). In the accounts that follow, I attempt to avoid “othering” or sensationalizing challenges that participants face in navigating the political landscape and underscore that these challenges are rooted in the media’s turn toward outrage and polarization rather than any kind of pathology endemic to the American heartland (e.g., Frank Reference Frank2004).
My identity as a researcher also raises questions about the extent to which participants may have chosen not to express antiexpertise tendencies to an academic hopeful. Based on my conversations, this seems somewhat unlikely. Over the year and a half that I knew them, conservative participants often told me stories about how universities “made” their friends or family liberal and one referred to a PhD as “piled high dogsh*t.” In other words, they were not shy about expressing anti-intellectual sentiments in general. Some even informed me that my outsider status—someone they perceived as neutral—afforded them a rare opportunity to speak about politics without fear of judgment. Others were more hesitant and occasionally offered comments such as “I don’t want to offend anyone;” however, they did so as a preface to articulating potentially offensive comments, including political opinions and community gossip. Although it’s impossible to know what participants chose not to share, I was also able to triangulate the findings through multiple interviews and lines of questioning, and I found very little difference within individuals over time.
Data Limitations
The data also have important limitations. The first has to do with representativeness. There may be differences between the kinds of people who are willing to participate in four interviews about politics and those who are not. As described above, I sought to mitigate this challenge by reaching out to people who usually avoid politics; however, attrition also tended to be concentrated among those who were apolitical. These losses indicate that the challenges of distrust and low political knowledge that I document here may be understating the experiences of my original sample.
Moreover, the data focus only on a sample of white people living in the small-town Midwest—a sample collected to answer a different research question. People with different racial or class identities have different relationships with authority in ways that shape their political trust (Hetherington and Globetti Reference Hetherington and Globetti2002; Wilkes Reference Wilkes2015; Wroe Reference Wroe2016), which limits my ability to draw population-level inference from these data. That said, participants’ demographic similarities helped isolate the variation that mattered for opinion formation. Second, while snowball sampling was useful for achieving a balanced sample, it also means that some participants know each other. This meant that I had to take care when analyzing data from participants in the same network to ensure that the conclusions I draw about the community do not rely too heavily on their observations.
Despite these limitations, the data retain important characteristics that make them suited to tracing distinctive processes of opinion formation by capturing the experiences of people with varying levels of political acumen as they searched for, gathered, and interpreted information about a novel issue outside of an experimental setting. That being said, I limit my conclusions to those that can be drawn from leveraging within-sample variation rather than extrapolating to other populations.
Findings
The argument proceeds in three parts. First, I describe how differences in media trust and political knowledge shaped participants’ prepandemic strategies for gathering information. Next, I show how this distrust led to distinctive pathways of opinion formation during the pandemic. I document rapid polarization among those with selective trust in partisan sources—affirming what surveys have shown (Allcott et al. Reference Allcott, Boxell, Conway, Gentzkow, Thaler and Yang2020)—alongside two different processes among participants with near-universal or universal distrust in all media. These processes both contributed to lack of opinion formation and political disengagement rather than polarization. And finally, I consider the role that alternative forms of trust, in government and in experts, may have played in producing this combination of polarization and lack of opinion.
Media Distrust and Information Gathering Prepandemic
An overwhelming number of people with whom I spoke in 2019 declared some degree of distrust in the “media.” This was true of Democrats, Republicans, and Independents. Their pervasive distrust was rooted in the fact that participants saw the media as a partisan institution, co-opted by political elites for their own ends. Despite this common ground, there was important variation in participants’ degrees of distrust, which shaped how they gathered information and derived opinions about government performance during the pandemic a year later.
Table 2 summarizes three categories of media distrust and political knowledge among my participants and the information gathering strategies that go along with each. Trusted sourcers are distinguished by the fact that they trust one, usually partisan, media outlet. Although they may not trust any other news organizations, this kind of selective (dis)trust means that the way they gather information is simple, relative to both triangulators and nonbelievers: it is a kind of one-stop shopping.
Table 2. Categories of Distrust and Information Gathering
Note: These categories point to where people fall on a grid of distrust and political knowledge, so some participants fell in between two categories.
Sophy, a Republican from Meriville, is one of these. Like many trusted sourcers, she claims to have a universal distrust of the media because they are all partisan; however, this belies her actual tendency toward selective exposure. As she told me when we first met in summer 2019,
I don’t understand how—you have CNN, you have Fox, you have MSNBC—how one news coverage can have a complete different perspective. It’s either the news or it’s not the news. Well, no. Then they–both parties, I’m sure, are guilty of this—they doctor it up. I mean, it’s like fake news. But, I think both sides are a little bit guilty of it.
Sophy’s concern about the difference between “news” and partisan “perspectives” is widely shared among my interviewees. Regardless of their own partisanship, participants agree that the media suffers from partisan biases, although only Republicans use the term “fake news.” But even as Sophy insists that all media outlets are suspect, she watches only Fox or local news.
In other words, for some trusted sourcers like Sophy, selective distrust is hidden beneath rhetoric of universal distrust. Art, for example, is an avid Fox News watcher—it was on in the background the first time we met in 2019. When I ask him a year later about how much he trusts the media when it comes to conveying information about COVID, he tells me, “I wouldn’t trust the media no further than I could throw ‘em.” But this is not actually true, as my own experience with Art indicates. I ask him about whether he trusts Fox, and he explains, “I just like this one. I think they come closer to telling the truth. I know that’s all you can go on, but at least they say what the other ones don’t.” In other words, they’re the best of a bad bunch.
But trusted sourcers share more than just a preference for the media they see as aligning with their politics; they also maintain a clear partisan map of the available news sources that allows them to selectively trust-in party sources and distrust out-party ones. Todd from Meriville, for example, notes some kind of bias to NPR but reserves most of his disdain for Fox: “My main source of news, just because—I know they’re liberal, but they’re fair—is NPR… . I do watch several different news sources. Sometimes I just need to feel mad, so I watch a little bit of Fox just to see what they’re blathering on about.”
As these examples illustrate, trusted sourcers often accept that all media have some partisan bias, but they settle for the source they see as least biased or in alignment with their partisan views. Even when they might articulate universal distrust that casts doubt on their own preferred media—like Sophy—or describe moments when they consult multiple news sources—like Todd—their information gathering practices are defined by one-stop shopping, enabled by sufficient trust in one or two sources. This, more so than political leanings, is what distinguishes their process of opinion formation from that of both triangulators and nonbelievers.
This is because triangulators come closer to universal distrust of media and have a somewhat less clear partisan map of the media to guide them. As a result, fact-checking has become a regular feature of acquiring and processing political information. Contrast Todd’s comments above to Danny’s, a Democrat from Iverson. The first time we meet in the summer of 2019, he tells me,
Danny: I isolate myself nowadays. For the past, I would say, four years, I’ve stepped away from social media. I stepped away from watching television, mainly because I refuse to pay so much for it. And then I’d say I just don’t … I don’t like how the media is. And it’s nothing new, I just, I like to come to my own conclusions. So if I’m doing things, it’s going to be based off of like other platforms like Reddit. And I’m going to look at it from eight different perspectives instead of just one.
Me: So you don’t watch the news anymore. Do you read any papers?
Danny: I don’t. I read Reddit and, I would say—yeah, primarily Reddit. I subscribe to, pretty much all the major news sources and then even some B.S. news sources. I try and drive my opinion from seeing it everywhere. But I take everything with a grain of salt. Like I really don’t believe anything I see. It’s more just, like, oh that happened today.
Danny, like other triangulators, maintains a more widespread, though not fully universal, distrust in the media such that he believes there is an inherent benefit to consuming political information “from eight different perspectives instead of just one.” Thus, triangulators often “do their own research” to get to the truth of the matter. Although Danny gets his news from Reddit, he doesn’t just read a comment or a link and take it for granted; rather, he corroborates those accounts of the news against others.
Also unlike trusted sourcers, their partisan maps can be somewhat amorphous. Danny, for example, makes the distinction between “major news sources” and “B.S. news sources,” but he explains that he takes everything with a grain of salt. Even so, triangulators usually have sufficient trust in certain sources and enough political knowledge to recognize, for example, when people offer partisan commentary on a Reddit thread.
Given these understandings and some minimum of trust, they can triangulate between different partisan sources. Often, they describe seeking some “average” by watching both CNN and Fox. Colton, for example, is an Independent from Iverson who tells me that he doesn’t trust either outlet. He continues, “I mean, each news source, they put their own spin on it. I guess … I try to groom through, figure out what the actual facts are and what they’re trying to sway me on.” In this process, a minimum of trust is essential. If Colton distrusted every piece of news completely, even with his partisan map of the media, the average of various partisan claims would still not provide a trustworthy foundation on which to base an opinion. Colton has enough trust to believe he can “figure out what the actual facts are” beneath the bias.
In effect, because triangulators have some minimum of trust and a sense of the partisan objectives (often perceived more than actual) of different media sources, they can weigh different claims, discount for partisan biases, and ultimately arrive at an opinion. Triangulators may also have sufficient political knowledge to connect these truths to the political world, although this varies. Danny, quoted above, takes for granted that the Democrats “really do want to change things and make a better economy.” Having some partisan map of the media, along with these understanding of that for which the parties stand, helps guide Danny’s conclusions about politics. These understandings stem not from the media but in large part from Danny’s experiences in a unionized, Democratic community.
The real challenges to acquiring new political information arise when people have a universal distrust in the media and also lack a partisan map to guide their news consumption. These are the nonbelievers. Abigail, a Republican from Meriville who would like to become more informed about politics, is an example of this group. Despite her desire to become an informed voter, she often expresses frustration about how to begin. This is because her lack of political information is exacerbated by the fact that she has no sense of which sources might be more trustworthy than others are. As she tells me at one point, “I don’t have a good news source. I really don’t and that’s bad.” As a result, Abigail often ends up opting out of information altogether. As she explained of her experience during the 2016 presidential election, she found it nearly impossible to evaluate Clinton and Trump as candidates based on what she saw on social media and in the news: “[How to] filter it, evaluate it, and analyze it? And is this really true? And should I believe what this celebrity is saying? And I don’t know. And why do I even read that or hear it?” For people like Abigail, the lack of a partisan map to the media makes it nearly impossible to evaluate the credibility of any piece of news.
Moreover, as Abigail’s case makes clear, universal distrust of media and low political knowledge unravel both components of political opinion formation. She often cannot form any opinion because she doesn’t know which news sources to trust, and this is particularly true when it comes to evaluating new issues or political candidates. But even when she does form an opinion—as she explains, she does so by talking to her husband and thinking through her own morals—she can also feel unsure how to match it to the parties. These challenges are indicative of those faced by other nonbelievers.
Media Distrust, Information Gathering, and Opinion Formation during the Pandemic
Regardless of which category they were in, participants’ distrust in media and political knowledge shaped their information gathering and processing strategies long before COVID-19. And despite the increasing spread of misinformation (Donovan and Wardle Reference Donovan and Wardle2020) and the complexity of COVID-19 issues, their distrust in media remained fairly stable through the pandemic, as did the way they gathered information in light of that distrust. But the novelty of the issue did pose increased challenges to opinion formation for both triangulators and nonbelievers.
Some triangulators, who could usually arrive at political opinions before the pandemic, began to couch their views in a language of uncertainty. It was difficult both to track the partisan debate about pandemic policy and to figure out any opinion among myriad sources. And for nonbelievers, forming any opinion about COVID-19, let alone connecting that opinion to formal politics, proved nearly impossible. Only trusted sourcers and a few triangulators developed a high degree of certainty in their partisan opinions about COVID-19 policies and found ways to justify them.
This combination of universal or near-universal distrust—the feeling of “I don’t know what to believe”—and selective distrust—those who believed only certain (usually partisan) news sources—both contributed to the same aggregate outcome within a few weeks of the first shutdowns. The tail ends of participants’ attitudes about pandemic policies and opinions about the severity of the virus were polarizing, but the “center” was not composed of those who fell at some midpoint along the issue position spectrum; it contained people who doubted their own ideas about the pandemic.
Polarization among Partisans with Trusted Sources
The first time I spoke with people during the early days of state-wide lockdowns in March, they often expressed a muddled, frantic view of the pandemic’s challenges. Rose, for example, is the Democrat from Iverson quoted in the introduction, whose first concern when we spoke in mid-March was the economic consequences of lockdowns: “Our life will be changed for a long time, especially here where a lot of the jobs are paycheck to paycheck. There are so many people now without jobs and it’s only going to get worse.” But in the next breath, she mentioned all the measures she and her family, who care for her elderly mother, were taking to protect themselves. “The last thing I want to do is expose my mother to anything,” she explained. Regardless of partisanship or where they lived, voters were well aware that the pandemic would have economic consequences—for their own jobs and the macroeconomy—as well as health consequences—for their personal well-being and the nation’s public health. Health and the economy were not in opposition.
But by April and May when I asked about their level of concern, partisans had begun to clearly express the view that there was a trade-off between these two metrics for handling the pandemic: the country could only mitigate one by exacerbating the other. As Christopher, a Democrat from Iverson, explained in mid-May, “In my personal opinion, I think there’s only two sides to this. You either want to help stop the disease. Or you don’t. And you want it to continue spreading. Even if—whatever your reasoning may be—that’s what you’re doing.”
Not only did these participants think in terms of trade-offs at this time; they were also thinking along partisan lines. Consistent with survey evidence (Allcott et al. Reference Allcott, Boxell, Conway, Gentzkow, Thaler and Yang2020), Republicans in my sample favored reopening sooner to protect the economy and Democrats favored extending stay-at-home orders longer to protect public health. In expressing their readiness for reopening in April and May, many Republicans shared the notion that “we can’t stay locked down forever.” Fred, for example, is a Republican from Meriville who described how he hoped Governor Holcomb of Indiana would approach the reopening question when we spoke in late April:
Well, I think the Governor has to talk to his experts and see what is going on in the state. And I think what he’s gonna do is do a partial opening of the state. And I don’t think that’s too soon. I think, the longer we wait, the more damage is gonna be done. You can only run so far on the gas tank when the needle’s on empty. And I think that’s where we’re runnin’ right now, as far as the economy is concerned.
Fred had told me the month before that he preferred the “scientific approach” to the pandemic. As is clear here, he still hopes his governor will listen to experts, but now he feels he must balance those concerns with economic considerations.
On the flip side, many Democrats like Jamie from Iverson hoped that the government’s response would prioritize public health. In mid-May, after Wisconsin’s supreme court summarily reopened the state, she told me, “If we could find ways to open up and do it safely, that’s fine. But this whole, ‘we must open up at the risk of everyone’s health, because the economy,’ is so dang stupid and short-sighted. You know who also doesn’t contribute to the economy? Dead people!” For Jamie and other Democrats, the health costs of reopening the economy could not be balanced by any financial gain. But what Jamie shares with Fred, and others with trusted sources, is that she recognizes the two opposing sides to the debate and plants herself firmly in line with her party.
Thus, despite the novelty of the issue, partisans with trusted sources quickly perceived and began to accept the terms of the reopening debate—as a trade-off between health and the economy—and simultaneously deduced which party supported each side. Moreover, the distance between partisans’ opinions only continued to diverge over the following months such that by the fall Republicans were living in a reality in which the pandemic was “overblown” and Democrats felt that Trump had offered one of the worst performances of any presidential administration in history.
As this happened, they leveraged various interpretations of the pandemic’s trajectory to justify their opinions. Some pointed to potential data manipulation, but more often, Democrats and Republicans agreed on the data—or on the fact that there were unavoidable inaccuracies in the data due to human error—but disagreed on the interpretation. As Jeff, a Republican from Williston told me in September 2020, “I think now that the evidence is clear that … It’s not as dangerous as what everybody led you to believe, that we’ve got to put people back to work again. And quit makin’ em so scared. We did a good job of making a lot of people scared.” Jeff is one of many Republicans who saw the rapid decline in COVID cases after spring 2020 lockdowns as evidence that early models of the pandemic’s trajectory were wrong.
But even when Democrats and Republicans disagreed on how to interpret the evidence, doing so did not undermine their ability to form political opinions. In fact, as Jeff’s case illustrates, participants often felt certain that the COVID evidence they relied on was “clear” precisely because they trusted information from certain news outlets.
Lack of Opinion for Those without Trusted Sources
But for others without a trusted source of news, the trajectory of opinion formation was far less linear. Some triangulators continued comparing and consulting multiple sources, which allowed them to form opinions and connect them to politics. Like trusted sourcers, they ultimately polarized. But no length of time was sufficient among other triangulators and nonbelievers, for whom forming an opinion and connecting it to politics proved challenging.
Among triangulators, doubts about the media’s trustworthiness continued to sow doubts about their own political opinions. Often, this was not because they couldn’t map opinions onto the political system, as they had sufficient knowledge to follow many partisan debates; rather, it was that they were too uncertain in their opinions to confidently land on either side of those debates.
Kyle, for example, is a Democratic-leaning Independent from Meriville who was planning to vote in his first presidential election in 2020. When I ask him how much he trusts the media’s information about COVID in October 2020, he explains,
I don’t know. It just varies. Media just plays such a big role in everything in the society that it’s hard to know what the truth is … so it’s very influential depending on which news source you go to … so you had one side pushing and said, “Ninety-three percent of these deaths weren’t even because of COVID,” whereas like the other side’s pushing like, “Ninety-three percent, they had underlying issues but they still wouldn’t have died if they didn’t have COVID for the most part.”
Kyle’s case is somewhat typical of triangulators at this time. He has an underlying trust that the data the media presents are accurate, but he is also aware of competing partisan interpretations of those data. He offers up the Republican view of COVID’s deadliness: that COVID deaths are overreported because they include people who are dying “with” COVID not “because of” COVID, as others explain. But he is also aware of the Democratic interpretation of those facts—those people would not have died without COVID. These were, in fact, the two prevailing partisan interpretations participants offered during fall 2020.
But although partisans with trusted sources could quickly bat down one or the other reading, triangulators and nonbelievers did not find this so simple. For example, contrast Kyle’s uncertainty with the confidence that Jeff, quoted above, offers after watching Fox News. For Jeff, the evidence offers a “clear” interpretation, but for Kyle it does not. In the end, Kyle tells me he probably favors the Democratic view. But who can be sure?
But nonbelievers fared even worse. With little sense of the partisan accounts of the pandemic or of which media sources articulated them, they both struggled to form an opinion and map it to the political system. Ben, an Independent from Williston, offers a relatively rare case in which distrust in the federal government exacerbated a near-universal distrust in the media. As he told me in April, “It’s hard to find places that will give you data rather than just give you their own opinion on what’s going on… . I have to take a look at what people say to see if what they’re saying is accurate or are they just trying to cash in on that?”
Every information source was suspect for Ben—he wished he could inspect the hospital records of COVID patients. Short of this, he struggled to figure out what the “truth” was as he watched the possible versions of reality multiply around him. As a result, he concluded in May, “I feel like I’m unable to comment on the coronavirus hardly.” While Ben’s conclusion might seem difficult to comprehend in light of how COVID-19 overwhelmed every aspect of daily life throughout 2020, it stems from a lack of trust in how the media—and in this case, the federal government—were conveying information about the virus and its spread. This feeling was deeply disconcerting among a substantial minority of people with which I spoke.
The result, for many nonbelievers, is that competing views on reality are generally considered to have equal claims to scientific validity. Figuring out which version is more believable is a time-consuming process, and many people just stop trying. Linda, for example, is a Republican from Meriville who tends to avoid politics. In October, she explains,
I guess, the problem is that people don’t necessarily go to the government to get their information. They’re getting their information from someone who has already like extrapolated the data, and this is the data story that they put together in their reporting, so finding those unbiased, [laughing] legitimate news sources. I’m not going to take the time to go to the CDC website myself if I know I have data, like I’m going to—and most people are not either.
For Linda, the possibility of finding the “unbiased” truth in the news is laughable. This means that she has to “do her own research,” as many people tell me, but this generally takes too much time for a working mother of four. Linda concludes by telling me that she is left in a similar kind of uncertainty to Kyle’s, but she also lacks a clear sense of where her partisanship might guide her in those debates. Thus, she tends to weigh alternative interpretations of reality equally.
From these responses, we can see how people might struggle to reach any conclusion about the pandemic and the policies designed to control it, let alone use those opinions as a basis for evaluating government officials or political candidates. The widespread sense that the media is constantly putting a spin on things left a number of interviewees feeling like they “don’t know what it means” when they are looking at case and death counts—resulting in a kind of “disorientation” (Rosenblum and Muirhead Reference Rosenblum and Muirhead2020). Moreover, the combination of media distrust and low political knowledge creates a formidable barrier to entry for nonbelievers. As Abigail, quoted above describing her difficulty evaluating Trump and Clinton as candidates in 2016, told me in fall 2020, “I haven’t done enough in looking at the things that I’m reading to know—Is this one-sided? Is this pro-Trump? Is this anti-Trump? Is this pro-Biden? Is this anti-Biden?” Lacking triangulators’ sense of how much weight to give to competing, partisan accounts, nonbelievers like Abigail struggle to evaluate many political objects. And although near-universal distrust of media is not limited to a particular age group—Ben and Kyle have just graduated high school and Abigail and Linda are middle-aged—for young people like Ben and Kyle, this means that it was difficult to even step a toe in the water of politics as they approached their first presidential election.
The resulting mix of polarization and lack of opinion formation was jarring. Partisans expressed staunch views of how the pandemic was handled that amounted to different versions of reality, whereas those without trusted sources remained uncertain about how to interpret basic facts, let alone which side of any debate they might fall on.
Other Kinds of Distrust
But as discussed above, people may have mitigated their distrust in media by turning to other information sources during the pandemic—namely, the government and public health experts. Or, as in Ben’s case, distrust in those sources may have exacerbated a lack of opinion formation. Given the increasing attacks by populists and conspiracists on the legitimacy of multiple sources of political information, this possibility seems likely. And yet, as this section will show, even where there was evidence of eroding trust in these sources, it generally did not contribute to a lack of opinion formation—except in a handful of cases, like Ben’s—as those who distrusted the government or disputed the experts tended to do so because they trusted information they gained from the media.
But even these cases were somewhat uncommon. Distrust in information sources outside of the media was fairly limited, particularly when it came to scientists and experts. Participants regularly told me that they expected their political leaders to implement policy based on expert guidance and described how they sought to follow that guidance themselves. As Fred, the retiree and Republican from Meriville quoted above, explained when we spoke about the virus in mid-March, “But I think what we’ve done so far has been correct. It’s an inconvenience, but still… . It’s been determined by people that are a lot smarter than me that social distancing is a good way to prevent its spread … and I think that the good, scientific, steady approach is the only way to handle it.” Fred was not alone in his desire to trust the experts. His stance, in fact, represents the overwhelming majority of people with whom I spoke, and it persisted up to the election. As participants formed their opinions about the government’s response to the virus and the nature of the crisis, they often expressed not just trust in the experts but a strong willingness to abide by how the experts told them to behave.
But this does not mean that everyone I spoke to shared the same understanding of what the scientific consensus was—in fact, by the fall there was a great deal of debate over the “correct” scientific approach to the pandemic. But in disputing current approaches, people often appealed to scientific expertise. Nate, for example, is a Republican from Williston who tells me in October 2020 that he’s not sure how he feels about pandemic policies because the science “keeps evolving.” As he explains, “Early on, they were finding out new stuff about coronavirus every day, right? The science on it will develop more and so we’ll know more over the coming years.” Until the scientists reach a consensus, he’s willing to defer judgment on the Trump Administration’s handling of the pandemic.
Nate was also among a substantial minority—usually Republicans—who offered up bunk science or called the early COVID models “wrong” (because the worst-case scenario had not happened) in order to refute containment measures. But even on these occasions, people made these claims to call politicians unscientific for not adapting containment measures to the current reality of the COVID data. In many ways, these appeals to science are unsurprising, given how “the participatory challenge to expertise” (Brubaker Reference Brubaker2021, 4) has allowed competing epistemic authorities to flourish.
But distrust in expertise matters here insofar as it is related to trajectories of political opinion formation. And for the most part, those who contested the data or argued the early COVID models were wrong did so in support of a clear political opinion, as described above.
But although there was widespread agreement that the scientific consensus—whatever that was—should reign, participants did disagree over the extent to which they could trust the federal government. Democrats became, by fall 2020, increasingly less willing to trust the CDC and other federal agencies. Christopher, for example, offers a typical response among a substantial minority of Democrats whose trust in the federal government was eroding: “Realistically, the Center for Infectious Diseases and the CDC should be the two things that we trust the most, but those have been politicized. We have seen yes men in the upper levels. Not as much in the Infectious Disease, because I think Fauci is a hero.”Footnote 8 Among these Democrats, trust in government was eroding as they saw “yes men” offering public health guidance to please the president rather than protect the public.
But even in this instance, distrust was rarely connected to opinion formation because it was those with trusted media sources who learned the details of the Trump Administration’s maneuvering within the CDC. And as their trust in the federal government eroded, they could still fall back on MSNBC, CNN, or the New York Times. When I asked Rose, quoted in the introduction, if she has greater trust in the government or media to tell the truth about COVID-19 case and death counts, she tells me, “I think the media tries to put a more honest slant on things.” But Rose, as described above, is a trusted sourcer who is referring only to ABC and CNN when she expresses trust in media.
And for the majority who retained trust in government, it was also rare that they used federal government information as a counterbalance to media distrust. Although some suggested this was a good strategy in principle, the time required to do so can be off-putting—as Linda from Meriville, quoted above, indicated.
In summary, the structure of the American media system and political assaults on its impartiality produce the distrust that underlies the distinctive processes of opinion formation I describe here. This is not to say that declining trust in other knowledge-producing institutions doesn’t have important political consequences, but among my sample of Midwesterners, these kinds of trust mattered less in terms of their ability to form political opinions. Often this was because participants’ ideas about how much they should trust in the government or experts actually stemmed from the media.
Discussion
The aggregate result of these distinctive processes of opinion formation was similar to what Levendusky (Reference Levendusky2013) finds in his study about the effects of partisan media on polarization: only the already-polarized moved further to the poles. And as in that case, only the well-informed partisans within my sample were willing to confidently speak their minds. This outcome, moreover, was not specific to COVID-19 issues but was the product of participants’ attempts to grapple with a novel issue using preexisting strategies for forming opinions in a low-trust information environment. That environment itself exists because politicians’ attacks on media impartiality and the existence of some partisan news have undermined trust in all news (Rosenblum and Muirhead Reference Rosenblum and Muirhead2020). The result is that many of my participants take for granted that everything they read or see is partisan.
The analyses presented here have also shown how issue polarization at the extreme tails of the distribution can hide important differences in the processes of opinion formation for those we might otherwise think of as at the “center.” Among the Midwesterners I spoke to, trust in media was a defining feature of whether or not they were able to form confident opinions and political knowledge was essential in guiding them to connect those opinions to politics. Although the most vocal among them were the most polarized, the least vocal were not necessarily centrist or moderate, but uncertain.
Moreover, this analysis offers some insight into differences between expressed distrust in media and actual media consumption practices. Trusted sourcers often claim, like Art, to harbor a hatred for all media, but in reality they consume only partisan news. As a result, they have little trouble arriving at confident—though usually polarized—political conclusions. Even triangulators can often rely on underlying trust in the facts, partisan maps of the media, and some knowledge of partisan debates to guide them to an opinion. Even so, this process is often time consuming and occasionally concludes in uncertainty.
It was only those with universal distrust in all media and low levels of political knowledge who populated the group of the truly uncertain. Among this group, conspiracy theories and different interpretations of scientific evidence are often given equal weight, as there is no way to differentiate one piece of information from another. Within my sample, this was the minority of participants. But because of the way that universal distrust presents a barrier to gaining more political knowledge, this may be an increasingly important political phenomenon amidst attempts to delegitimize the media, government, and expertise.
And finally, my data suggest that this group may be politically important, even if they are a minority. Participants with the least confident opinions about COVID-19 were also some of those least likely to know for whom they were voting up until the very last moment. With an increasingly small portion of the electorate up for grabs as “swing voters,” understanding what drives their thinking about political objects—in this case, uncertainty—is increasingly important.
That being said, this analysis cannot assess how these processes of opinion formation play out in the population writ large, in other demographic subgroups outside of white, small-town Midwesterners, or with respect to other issues. Future research can take up these questions using both qualitative and quantitative methods.
Interview studies are well suited to exploring the relative contribution of media distrust to political opinion formation within differently racialized groups. For example, Black Americans have deep-seated reasons to distrust the government given centuries of state-led oppression (Wilkes Reference Wilkes2015). This suggests that experiences like Ben’s from Williston—in which distrust in government compounds distrust in media—may be more salient among Black than white Americans. A comparative study can shed light on these questions.
The argument presented here also suggests a fourth category of social actors that did not appear in these data—those with universal distrust in media and high levels of political knowledge. The fact that universal distrust and low political knowledge are self-reinforcing may be one indication as to why my sample did not include this category, despite the fact that it likely does exist among people who learn about the political world from within their social networks. It is not clear how this would affect the process of opinion formation. These people may take partisan cues from friends and family and let their partisanship guide them when they consider new issues, or their distrust in media may make them discount how others arrived at opinions and leave them feeling as uncertain as nonbelievers. Future interview studies might sample selectively on people with high political knowledge to explore how varying degrees of trust in media affect the processes of opinion formation within this group.
But nationally representative surveys are necessary to understand the prevalence of the relationships documented here. And although various surveys already track trust in media and government, both in the US and cross-nationally, my analyses suggest they may need to probe expressions of distrust more carefully to capture the full range of its political consequences. In particular, my conversations with trusted sourcers suggest that many of the people who claim to distrust the media—particularly conservatives who are comfortable using the term “fake news”—are actually distinguished by the very fact that they do trust some media (but see Jerolmack and Khan Reference Jerolmack and Khan2014). Surveys could more explicitly take into account differences between talk and practice to capture this phenomenon.
And finally, COVID-19 was an exceptional issue. It was not only new but also uncertain, and it had grave effects on Americans’ daily lives. In examining how trust factors into opinion formation, future research should evaluate how Americans gather and process information about new and “old” issues, along with issues that have less of an immediate effect on their lives. But the evidence from my participants suggests that when an issue is less grave, threatening, or all-consuming, Americans who face the double barrier of low political knowledge and universal distrust of media will be even more likely to avoid expending the extraordinary effort required to form a political opinion.
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://doi.org/10.1017/S000305542200003X.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/TVCR8J.
ACKNOWLEDGMENTS
Thank you as always to Lis Clemens and Eric Oliver for reading and commenting on my work. Thank you also to Geneva Cole and to Anna Berg, Nisarg Mehta, Nicolás Torres-Echeverry, and all the participants in the Great Mediation Conference.
FUNDING STATEMENT
This research was funded by the National Science Foundation (Award No. 2001815) and the Social Science Research Council’s Social Data Initiative.
CONFLICT OF INTEREST
The author declares no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The author declares the human subjects research in this article was reviewed and approved by the Institutional Review Board at the University of Chicago and certificate numbers are provided in the text or appendix. The author affirms that this article adheres to the APSA’s “Principles and Guidance on Human Subject Research.”
Open access
