Recent U.S. history has shown the challenges our country faces translating an increasingly divided and hostile political culture into effective governance. Misinformation, or “fake news,” is often cited as a central cause of this strain due to its perceived power to blur the right course of action and drive Americans towards extreme political views or even acts of violence. The spread of misinformation has the potential to undermine our democracy—resulting in more extremists gaining power or malicious actors deceiving voters in ways that push the country toward ends that benefit only them.

However, while the power of misinformation is often taken as self-evident, it is less clear why it has become such a dominant force in our politics. Existing explanations tend to focus on the 24-hour news cycle, partisan news networks, the viral nature of social media and the echo chambers it creates, low levels of media literacy, and an absence of fact-checking or verification mechanisms, among many other factors. The rise of digital social media, in particular, is blamed for much of the prevalence of misinformation today.

But all of these narratives limit our ability to understand the problem. To have an effect, misinformation needs to be not just present in our media environment, but also attractive, useful, and beneficial to us as consumers. For some political actors who would use fake news to fool people for some ulterior motive, the benefit is clear. What is less clear is why misinformation might be appealing for the much larger portion of voters, who must be influenced by and willing to share it, for misinformation to have any substantial effect.

Drawing on recent research illustrating the many factors that determine our appetite for misinformation, this report makes the case that we need to consider our winner-take-all electoral system when we talk about misinformation. In particular, making the connection between winner-take-all elections, affective polarization, and misinformation can help spur structural solutions that weaken incentives to share and believe misinformation in the first place, and in doing so help drain the reservoir of misinformation instead of just plugging holes in the dam.

This report highlights the need for more research on how different voting methods shape incentives to create, share, and believe misinformation, and whether electoral reforms can weaken these incentives. It begins by explaining why misinformation is a problem for democratic systems and then reviews existing solutions for misinformation, clarifying why they fall short in fully accounting for the sources of misinformation. It then argues that we need to think about electoral systems when talking about misinformation, particularly because of how electoral systems can shape identity-based factors that increase susceptibility to misinformation. The report describes existing research that connects the winner-take-all system with heightened affective polarization and affective polarization with a greater likelihood of believing and sharing misinformation. It concludes by identifying areas that need more research to move our understanding of electoral systems and misinformation forward and recommends fruitful avenues for research.

Misinformation—defined here as deliberately shared wrong information (e.g., fake news), which is distinct from simple misperceptions (e.g., genuine mistakes)—poses a number of dangers to a democratic society. Democracy thrives on the active and honest participation of citizens and misinformation threatens its success by obfuscating or discouraging the best course of action for voters and distorting perceptions of political opponents.

Most obviously, misinformation can decrease the chances that people are voting in their real interests, or what the literature refers to as “correct voting.” One recent example of possible incorrect voting was the propensity of uninsured conservative Americans to oppose candidates who supported the Affordable Care Act due to misinformation about its policies and use of “death panels.”¹ Supplied with perfectly reliable information—unfiltered by partisanship—many of these same voters might have supported the legislation that would have given them affordable healthcare, and thus had a real benefit for their lives.

Incorrect voting like this is an especially pernicious possible effect of misinformation. Because a democracy relies on voting, the entire apparatus of government may lose legitimacy as a result of too many voters being fooled into supporting candidates, parties, or policies that do not actually benefit them. This should be especially worrying in the American context where—due to the relatively large number of candidates, ballot measures, offices, flexibility of the platforms of major parties, and frequency of elections—voters must invest significant time and energy to obtain accurate information, making correct voting difficult even without the presence of misinformation.²

Misinformation can also reduce political participation by clouding the truth and sowing distrust in both the infrastructure of government and political actors. One channel through which misinformation could do this is by distorting perceptions of procedural fairness in how the government and elections work. Voters’ ideas of the procedural fairness of a governmental action—their perception that a system or process is objectively fair for all parties involved regardless of whether their party won—can have a significant impact on whether they think proceedings are legitimate.³ Procedural fairness can also be a significant measure of voters’ views on how their democracy is performing, and some research has even found that perceptions of fairness—including the role of money in politics and representation of women—far outweigh traditional “pocket book” economic concerns, which have long been thought to dominate politics.⁴ If misinformation is able to cast doubt on the fairness of a result, there is ample reason to be concerned this could impact voters’ perceptions of democratic legitimacy, and, as a result, their willingness to participate in or accept the result of elections.

Finally, if misinformation is used to attack political opponents, it may increase political hostility and extremism. While some level of animosity and zealotry are unavoidable in a democracy, the danger for a well-functioning democracy when these forces become extreme lie in their power to sow distrust and dissatisfaction with the system as a whole. A democracy relies on the consent of its losers. If political opponents are so extreme or despised as to constitute an existential threat to “our side” their victory, regardless of its root in a fair election, it will be perceived as an intolerable cataclysm. Existing research suggests that partisans on the losing side have become increasingly dissatisfied with democracy.⁵

While there are many commonly discussed solutions to misinformation, most focus on controlling the media environment as opposed to reducing the power and appeal of misinformation itself. All of these suggested solutions also presuppose some level of novelty to our current information diet, particularly focusing on social media as the greatest market for misinformation. Inherently, such narratives suggest that there was an age of politics in which voters had bountiful correct information and were not susceptible to misinformation. However, these ideas do not reflect the reality of the past or present state of U.S. politics: Misinformation, particularly in the form of paranoia and conspiracy theories, has been a durable feature of American politics for most of our history—a feature social media has amplified but certainly did not create.

In his 1964 article, “The Paranoid Style in American Politics,” Richard Hofstadter traces the consistent willingness to believe a wide variety of conspiracy theories among the American public. Hofstadter characterized the time as being dominated by “angry minds at work mainly among extreme right-wingers.”⁶ He cites as examples the widespread nature of baseless conspiracies against the Freemasons in the late eighteenth and early nineteenth centuries, Catholics and Jesuits in the latter half of the nineteenth century, and Communists during the McCarthyism of Hofstadter’s own era. While the lessons Hofstadter draws from this genealogy of conspiracy were meant to address the extremity of the mid-twentieth century, many are prescient of our current era of inter-party hostility and conspiracy mongering, illustrating more evidently than any historical record the durability of this issue in American politics.

Misinformation has been salient in our politics for many years, as shown by an analysis of National Opinion Research surveys regarding Americans’ beliefs about the assassination of John F. Kennedy.⁷ By the late 1960s, conspiracies regarding the assassination were already widely believed, reflecting America’s longtime willingness to believe and spread misinformation regardless of the media environment. Historical work like this can illustrate the salience of our infatuation with misinformation regardless of the media platform. Contemporaneous research on our current media environment only confirms this narrative, and clearly illustrates that while social media is an important factor in the spread of misinformation, it is not a root cause in itself.

Fixing Social Media?

One of the more popular responses to misinformation focuses on regulating social media. If social media is inherently given to misinformation or something about its use ensures people will believe or spread it, then this may be effective. However, if social media simply reflects trends that are being produced elsewhere, the effect of regulations will be severely limited. As new collaborative research between Meta and academics found, tweaking various aspects of Facebook and Instagram during the 2020 U.S. elections—like removing reshared content⁸ or providing a chronological news feed instead of a feed curated by the platforms’ algorithms⁹—did not significantly alter polarization or political attitudes.

As a variety of research indicates, social media seems to be amplifying existing behaviors. The conventional wisdom is that online interactions make people more hostile because in social media platforms we lack facial cues, have much more power to keep ourselves anonymous, and have more public discussions accessible by a great number of outside people. Essentially, our self-regulation or empathy may be weaker online because our opportunity to create conflict is greater while social cues that might normally discourage it are absent. However, these expectations are not supported by evidence. Research looking into various studies across different countries found that online and offline hostility are closely connected.¹⁰ Other work finds that, while the norms that moderate our behavior may very well be weaker online, the root cause is likely a dangerous lack of moderating norms for political discussions everywhere in our society, suggesting that our political elites, campaigns, and culture encourage hostility towards opposition, which social media may amplify but does not create.¹¹

Online political hostility may also be a way to signal status, reflecting deliberate decisions motivated by status-driven individuals, whose actions are much more visible online than they are offline.¹² Political hostility online is born from a social need to demonstrate one’s own allegiances and gain status. While the nature of online discussion makes such demonstrations more visible, their cause is not fundamentally different or separate from the cause of similar actions offline. Although these status-driven individuals using hostility online may be a minority compared to the backdrop of largely benign interactions, their activity can make online environments significantly more hostile and help spread misinformation more quickly.¹³

This evidence should not be understood to suggest that nothing has changed in the universe of political information. New media types and communication methods have made the distribution of information cheaper than ever, allowing those with incentives to share misinformation the ability to reach a greater audience in a less filtered manner.¹⁴ However, this broadly fits with a historical pattern: New media environments that result from the advent of new technology or social shifts (radio, mass newspaper, 24-hour news, and social media) create more opportunity for information sharing and thus more opportunity for misinformation to spread. The effect of different types of media is also heterogeneous, with different methods of delivering information having different or opposite effects on consumers’ engagement, feelings on democracy, and likelihood to vote or take other political action.¹⁵ While these conclusions do point to the importance of the method of information delivery, they also show that the underlying culture necessary for the widespread belief and spread of misinformation has already been present in our political culture for much longer than social media. Further, this illustrates that as long as the social incentives to share misinformation or become politically hostile exist in wider society, they will exist wherever social interactions happen regardless of regulation.

Diversifying Media Environments?

Going by conventional wisdom again, misinformation is amplified because Americans are largely siloed in their media consumption, especially when it comes to media found online or on social sites.¹⁶ Thus, to prevent false claims bouncing around in echo chambers—and being strengthened by repetition and lack of any challenge—we must diversify our media environment and break the average American out of their bubble. However, it turns out that the vast majority of Americans are far less isolated from dissenting political information online than we often think.

Mathew Gentzkow and Jesse Shapiro use aggregate website audience, individual level browsing, and Cross-National Election Study survey data to construct an index that captures how segregated media platforms are by ideology. Their index ranges from zero to 100, where zero means people from across the ideological spectrum obtain their news from the same sources, and 100 means that there is no overlap in the media sources across ideological groups.¹⁷

On this scale, the American internet had a score of 7.5, slightly above most traditional media sources which ranged from 1.8 (broadcast television) to 10.5 (national newspapers) but significantly below all offline social interactions which ranged from 14.5 (voluntary associations) to 39.4 (political discussions). These results lead Gentzkow and Shapiro to conclude that online information is less segregated than often thought, especially compared to our interpersonal interactions offline.

Other research further shows that face-to-face social interactions are far more segregated than online interactions.¹⁸ Social media, particularly non-anonymous social media, can lead to greater cross-cutting exposure than offline behavior because there is a higher level of multidimensionality as well as weaker norms discouraging disagreement in online social connections.¹⁹ Thus, while social media may not expose us to people who differ from us dramatically, it does expose us to a greater diversity of people than our other interactions and we may be more willing to discuss uncomfortable topics—like politics—in those online settings.

A more targeted investigation of Facebook user data reached similar conclusions, finding that about 20 percent of the average American’s friend network identifies with the opposite party.²⁰ This work also finds that, while there is a small chance they will interact with or share this content, users see a relatively large proportion of cross-cutting news on their newsfeeds (due to both the content curated by Facebook and what is shared by their network). The same methodology has been used to show that the average American’s media intake online is overwhelmingly moderate.²¹ On Twitter, while the majority of information shared was with those who were ideologically similar, users were exposed to a wide variety of sources of information depending on the subject matter.²² However, new research produced collaboratively between academics and Meta recently found that Facebook is more ideologically segregated than previously thought, especially when it comes to news shared in Facebook Pages and Groups compared to content shared by friends.²³ The problem with echo chambers, if they exist, thus has far less to do with the information we are exposed to than the information we are willing to share and make use of—the motivations for which are driven by social-identity factors—as established in the previous section.

There are two important caveats to this conclusion. First, the majority of Americans simply do not spend that much of their time consuming political news online, instead having only a casual interest in the subject.²⁴ Furthermore, those that are especially active in politics drive a disproportionate amount of the traffic towards partisan outlets.²⁵ So while the majority of Americans do not exist in media bubbles, a vocal and highly active minority might. Researchers looking at the efforts of the Russian Internet Research Agency (IRA) observed that those that are more likely to engage with its propaganda are already the most polarized, active on social media, and interested in politics.²⁶ This may in fact limit the damage caused by actors like the IRA, as their efforts primarily interact with the few Americans who are siloed online and reinforces the idea that those who want to share misinformation for political gain will find a way to do so.

Second, while overall information consumption is moderate in nature, Republicans, particularly Trump supporters, face greater exposure to misinformation while Democrats are more likely to share information that comes from across the aisle.²⁷ This is at least partly due to the supply of information on each side of the spectrum.²⁸ While the incentives to spread misinformation or information that is beneficial to one's own party may be the same for liberals and conservatives, the well of fake news may be deeper on the conservative side and take more work to avoid while still finding information useful to one’s partisan preferences. This is the case for Facebook, where conservatives consume more homogenous news content than liberals and most misinformation circulates in conservative networks on Facebook.²⁹

Fact-Checking?

Failing system-wide regulation and a paradigm shift in how Americans get their information, fact-checking is often pointed to as one of the easier and already practiced means to address misinformation. Inherently, this suggestion relies on the reason for misinformation’s salience being its believability—it assumes that if people were only shown that a narrative is a lie they will stop spreading, using, or believing it.

However, the power of social identity goals to create incentives to share misinformation points to an important takeaway for countering it: Fact-checking or simply correcting false information will not work as it does not counteract the social identity determinants of misinformation sharing. While it might clearly prove that a story is fake, fact-checking will not stop the spread of misinformation if the need to signal one’s politics, derogate the opposition, or generate chaos is a more powerful motivator than truth.³⁰ An experiment in Côte d’Ivoire tested this hypothesis, concluding that interventions aimed at reducing the belief and spread of misinformation were successful when they targeted social-identity factors, like empathy, social norms, and elite endorsements, but not when they promoted digital literacy.³¹

“Fact-checking will not stop the spread of misinformation if the need to signal one’s politics, derogate the opposition, or generate chaos is a more powerful motivator than truth.”

Indeed, fact-checking may often be more harmful than beneficial, as multiple studies have proven that even minimal corrections to flagrantly untrue stories can have a strong exposure effect: just having seen the correction makes people more likely to remember the false information.³² To have any hope of actually correcting misinformation, fact-checking must come from an unexpected messenger—someone from the in-group the misinformation benefits who is going against their own interests to set the record straight.³³ In contrast, correction from someone of the opposing party can cause acceptance of misinformation to grow stronger. Thus PolitiFact, Reuters, or any of the myriad of other fact-checking sites will almost always fail to counteract a lie told by Donald Trump because they are not a trusted member of the in-group sharing that information. To be effective, a correction would have to come from Trump himself or someone in his camp with enough credibility to go against the grain and set the record straight.

This is not to say that there is no intervention that might reduce misinformation’s power and appeal. As the Côte d’Ivoire study shows, interventions that target the motivations to use misinformation can be much more successful at curbing its use.³⁴

The common approaches to misinformation discussed earlier inadequately confront its threat because they overlook the pre-existing social factors that fuel its demand. Attempts to combat misinformation by altering the media landscape without concurrent efforts to address the political culture in which it exists are doomed to be ineffective at worst, or band-aids at best. Such strategies disregard the political system that incentivizes the production, dissemination, and acceptance of misinformation in the first place. By considering the effects of the political system—particularly of the winner-take-all electoral system on affective polarization—a narrative quickly takes shape linking our elections to the drivers of misinformation sharing, namely the need to signal in-party loyalty, denigrate opponents, and win at any cost.

When thinking about the causes of misinformation—and possible solutions to the problem—we need to think about our electoral system and how it shapes demand for misinformation through its effects on affective polarization, a measure of voters’ identification with their own party and dislike for out-party members. Our winner-take-all system creates strong incentives to dislike and fear political opponents and raises the stakes of winning a race, thus setting the stakes, playing field, and social mores in which a large part of our social and political identities develop. Our majoritarian first-past-the-post election method—which generally pits one party against the other in winner-take-all fights where it is easy to cast opponents as evil—contrasts with consensus-based systems of which the most common are proportional representation. The key metric here seems to be the degree to which the electoral system in question creates stronger ties between one's social identity and their politics, as well as the degree these ties become oppositional.

In an oppositional and winner-take-all electoral system, elected officials who cannot compromise on the issues most important to their constituents face increased incentives to obstruct policy changes and put the blame for inaction on their opponents.³⁵ In a coalition government these same actors would likely face a more difficult time blaming other parties as well as less incentive to do so because, as discussed later in this section, partisans at all levels would have less of an us-versus-them mentality.

Elite polarization along cultural lines has occurred in tandem with the sorting of parties along racial, religious, and identity-based divides since the 1970s. Analysis of the American National Election Survey (ANES) from 1972 to 2004 reveals that members of a party who are highly aligned with their party along social identity lines are less equipped to deal with a threat to their party and more likely to feel stronger negative emotions towards the out-party.³⁶ Perceptions of political opponents as a threat—and the associated hostility towards them—will almost certainly be more likely in a winner-take-all system that ensures that one is either part of the winning party or the losing one, as well as ensuring that being in the losing party means one's cultural identity is also losing.³⁷ ANES data also showed that in 2016, despite the historic unpopularity of both candidates among their own party’s supporters, partisans overwhelmingly voted for their party’s candidate due in large part to their outright hatred for the opposing party’s candidate.³⁸

Thus, as social identities have become increasingly wrapped up in party politics, the actions of the opposite party and the possibility that its representatives could win complete control of the government have motivated out-party animus. Without such a great risk that our ‘enemies’ could defeat us, this animus would likely decrease.

Our partisan politics have also created incentives to act based on animosity. Research using data from the 2010 Cooperative Congressional Election Study (CCES) finds that partisans are driven to act based on their perception of inter-party rivalry and anger with the opposition—each of which is driven by the strength of partisan attachment.³⁹ Moreover, negative stereotyping of opposing partisans as lacking moral traits (itself a consequence of affective partisanship and misinformation) is integral to creating this rivalrous relationship and anger.⁴⁰ All of these effects are exacerbated by close elections, distracting voters from the issues and towards identity-based conflict. As noted by the researchers, a stable two-party system is key to allowing partisans to identify both with their own party and classify the opposite party as an enemy.

Losing in a winner-take-all system is much more costly than in proportional systems and threatens the identity of partisans much more. Indeed, strongly identified partisans will react to the prospect of winning or losing (which acts as a threat to their core identity) with stronger emotions—particularly anger—than nonpartisans.⁴¹ Out-party hatred can even result in a positive reaction to objectively harmful events, such as a downturn in the national economy or the deaths of U.S. soldiers abroad, so long as it reflects poorly on the opposition party.⁴² Other work adds more evidence that strong emotions can cause hostility online, especially moral outrage.⁴³

Experimental tests of the effects of electoral systems on intergroup animosity provide further evidence that the winner-take-all system generates more hostility and distance between partisans. In dictator games played in contexts with more parties, for example, the affective gap was lowered.⁴⁴ Comparing similar dictator game-based tests of affective polarization confirms that the directly oppositional nature of a two-party system is key to assessing the strength of these experiments. In other research, dictator game participants showed consistent in-group favoritism and out-group penalties.⁴⁵ However, these and other such sociological tests of affective feeling do not separate in-group favoritism from out-group hatred because they are zero-sum: for one player to profit another must lose out.⁴⁶ In other experiments that are not zero-sum or that differentiate rewarding one group from actively punishing another, out-group punishment does not appear. Because first-past-the-post is winner-take-all in nature, it mirrors the simpler version of the dictator game used to model both in-group favoritism and out-group punishment in a zero-sum environment. A consensus-based system on the other hand, where multiple groups can share in victory and a win for me is not necessarily a loss for you, is closer to the alternate setup where preferences for one party do not translate to punishment for the other. In essence, first-past-the-post creates stronger and more negative affective polarization because it ensures that the ‘game’ must involve the punishment of one's opponents.

An analysis of 20 Western democracies offers further reason to investigate our elections, finding no significant relationship between electoral system and affective polarization, but rather a strong link between election method and what affective polarization represents.⁴⁷ Affective polarization—as the following section further elaborates—is usually constructed as the difference between in-group and out-group ratings. As shown in the chart reproduced below, proportional representation increases in-party identification and decreases out-party derogation while winner-take-all decreases in-party identification and increases out-party derogation. While the overall affective polarization score may not significantly differ between systems, in proportional representation systems this number represents a far “gentler and kinder” method of politics than in first-past-the-post.⁴⁸ These findings correspond strongly to the idea that proportional representation, by being less oppositional than winner-take-all, creates a less rivalrous, angry, and hostile politics.

One of the most pernicious consequences of the winner-take-all system is that it intensifies affective polarization. The simplest definition for affective polarization is the degree to which a partisan identifies with their own party and against an opposing party based on their social identity.⁴⁹ The inclusion of social identity is key here as affective polarization is distinct from, and possibly much more impactful than, ideological polarization—a simpler measure of the distance between partisans on policy.

There is evidence that affective polarization is on the rise using the most common measure: in-group vs. out-group thermometer differentials (the difference between how positively a voter feels about their own party compared to how positively they feel about another party). Data from six different surveys of voters in the United States and the United Kingdom show that while both major parties in the U.S. have a fairly consistent level of in-group support (positive feeling towards their own party) there is a clear downward trend in out-group feelings: While positive own-party ratings for both Democrats and Republicans remained relatively constant at around 70 percent between 1975 and 2010, positive feelings towards the opposing party dropped by almost 20 percent for both parties in the same period.⁵⁰ This increasing negative feeling does not appear to be related to specific elections or ideological battles. In fact, this divide of partisan identity is growing far stronger than the same cleavage between Black and white or Protestant and Catholic Americans.⁵¹ This fits with other findings that show that 70 percent of partisans show a consistent bias in favor of their own party on implicit bias tests and that partisan strength consistently had a stronger positive effect than ideological strength on the bias between parties, political activism, and anger felt towards the opposite party’s candidate.⁵²

Analysis of the Comparative Study of Electoral Systems (CSES) across multiple elections in 20 Western democracies reveals that Americans between 1996 and 2017 were relatively tepid towards their own parties, unusually hostile towards their opponents, and above average in terms of overall affective polarization, as shown in Figures 3 and 4.⁵³ This data also shows that out-party dislike has risen in the U.S. to a greater extent than in other countries observed.

In behavioral tests, Americans consistently reward co-partisans over members of the opposition in decisions about who to hire, how to award financial compensation, and who to award academic scholarships to.⁵⁴ Partisans also increasingly assign positive traits to their own party while negatively stereotyping the opposition as ‘selfish’ or ‘close-minded,’ among other such traits.⁵⁵ Americans have also become increasingly unhappy to see their child marry someone from the opposite party and are unlikely to date, marry, or even befriend someone from the opposite party themselves.⁵⁶

Thus, affective polarization represents the convergence of a political and social divide that Americans are increasingly unwilling to cross.⁵⁷ Emerging research shows that this growing affective hostility is a key cause of misinformation—and likely a much stronger one than the traditional explanation that Americans, particularly older Americans, are being fooled.⁵⁸ Instead, regardless of whether people believe it, misinformation is acted upon and shared due to partisan affiliations, the need to signal loyalty to a group, and, most importantly, dislike for opponents.

“Misinformation is acted upon and shared due to partisan affiliations, the need to signal loyalty to a group, and, most importantly, dislike for opponents.”

Cognitive psychology research tells us that people engage with or share information based on one of two orientations.⁵⁹ The first is their accuracy-oriented motivation, where they will engage with and accept true and reliable information to assist in the creation of an accurate worldview. The second is their goal-oriented motivation, which causes them to focus on information that is useful to their beliefs or allegiances. By agreeing with and spreading information that may not be true but is accepted by or beneficial to their social group, people are demonstrating their solidarity to the shared reality of that in-group (in this case their party).⁶⁰

It is tempting to imagine that this sort of goal-oriented behavior might be in service to ideological ends—in other words, that misinformation is a tool to win policy-based debates and ensure one's preferred course of action is taken. However, these actions more likely reveal social motivations to score points against an opposing party.⁶¹ Indeed, research finds that affective polarization, particularly negative animus towards political opponents, is central to predicting misinformation sharing.

Affective polarization can motivate misinformation sharing as a way to increase a person’s social status: People are much more likely to share negative or derogatory articles about the opposite party than about their own (64 percent compared to 25 percent).⁶² Thus, people sharing misinformation may be very able to discern fake from real news, but their need to signal their commitment to or promote the interest of an identity group (such as a political party) will win out over their motivation to be correct. This fits well with findings from a variety of researchers that the sharing of misinformation is not accidental, but an intentional decision to engage with political discussion and, sometimes, offend others both online and offline.⁶³ Libby Jenke further confirms this connection by exposing partisans to negative behavior from a perceived political opponent before testing their adherence to misinformation favoring their own party or the opponents. Participants exposed to negative behavior were significantly more likely to subscribe to misinformation that benefited them and less likely to subscribe to misinformation that benefited the opposition.⁶⁴

This is all not to say that misinformation is never believed or that that belief is not harmful, just that the main reason for it being shared or acted upon is not that it is erroneously believed due to a lack of knowledge or critical thinking ability.

The importance of affective polarization to misinformation suggests a worrying connection: If hostility towards our political opponents encourages us to share lies, and those lies then encourage further hostility, we will be trapped in a doom loop.⁶⁵ More plainly, when confronted with negativity about ourselves or our co-partisans, our negative response becomes even stronger as we circle the wagons. On the flip side, when repeatedly exposed to negativity about our opponents from within our own community, we are incentivized to display loyalty to the partisan beliefs about one’s opponents and in turn spread further distortions of the truth.

This feedback loop may be most apparent when we consider its implications for the behavior of elites. Elite actors, largely meaning here elected officials and other highly influential political figures (notable media pundits, candidates, community leaders, etc.), have a large and obvious effect on misinformation via their own sharing. Elites have the ability to share misinformation to a much wider audience than the average American as well as a much higher chance for that information to be taken seriously (or taken as the required narrative for social signaling purposes). Misinformation has a feedback effect on elected officials because they too must signal their allegiance to the narrative endorsed by their constituency. Thus, if a piece of misinformation is accepted by a given political identity, the elected officials representing that identity will face costs from their constituents if they do not signal their acceptance of that information regardless of its veracity. A feedback loop emerges due to the reach of elites who are incentivized to amplify misinformation by the same social identity goals that trump accuracy goals for normal individuals. The ideological extremity of elites of the opposite party to a voter drives this even more, as partisans respond by holding stronger feelings of partisan animosity.⁶⁶

Research using data on roll-call votes and the ANES survey from 1978 and 2016 finds that elite ideological polarization has increased and that this, combined with individual interest in politics, correlates with a rise in affective polarization at the popular level, with the ideological polarization of out-party elites having the largest effect.⁶⁷ Data from the Comparative Manifesto Project on 20 Western democracies, including the U.S., shows that elite ideological polarization around cultural issues (way of life, law and order, multiculturalism, etc.) is a significant driver of affective polarization whereas similar polarization on economic issues is not.⁶⁸ The difference in effect between the two types of elite polarization is likely due to the difficulty to compromise on principled moral issues compared to more pragmatic economic ones. Intense ideological disagreement in the legislature thus causes not ideological strife, but feeds into negative inter-party feelings that are wholly separate from ideology.

The salience of disagreements over cultural issues will likewise hold more generally for everyday partisans regardless of elites. Because these issues are more wrapped up in our social identities, challenges to them make it easy to make those we disagree with into our enemy or otherize them as irreconcilably different.⁶⁹ The focus of U.S. media on dramatic political conflict—and omission of examples of compromise—may further this othering, as it makes it easier to caricature the opposition as extreme and unreasonable.⁷⁰ Indeed, when exposed to negative online hostility towards members of their own in-group, both parties’ partisans registered higher affective partisanship as measured by the bias in their ratings of parties (though this result was stronger for Republicans).⁷¹All of this will make it more tempting, believable, or useful to share further misinformation demonizing political opponents. This feedback loop can be counteracted to some extent by correcting characterizations of our opponents and showing that political opponents are much more similar than they are different.⁷² However, doing so faces the same challenges highlighted earlier regarding fact-checking.

The largely institutional basis of the issues discussed in this paper beg the question: Why are misinformation and affective partisanship worsening now relative to any other time in the nation’s history (or across the world)? Furthermore, if our particular electoral system makes political hostility and misinformation worse, what accounts for the high levels of both in democracies that use consensus-based systems? Brazil, for instance, uses a proportional system but has high levels of inter-party hostility and fake news.

The institutions that govern how the government works have, after all, changed very little in over 200 years. The connection of misinformation with our electoral system that we have laid out does not mean only those with a first-past-the-post system will experience misinformation, just that they may be more likely to because such elections create a zero-sum competition between “us” and “them.” While we would expect to see this less in countries with more consensus-based governments, we cannot rule out the development of high affective polarization in such systems along different cleavages (establishment vs. anti-establishment, for instance), which would then lead to the use of misinformation. Further, we can offer two appealing narratives based on the evidence that might explain some of the particular difficulties in the United States.

First, the sharp growth in own-party versus out-party antagonism has coincided with the sorting of parties along racial, religious, and ideological lines.⁷³ This has ensured partisans react much more strongly to the perceived threat of opponents and diminished the traditional importance of candidates cultivating personal images tailored to their constituencies—as antagonistic partisanship provides little opportunity for differentiation from the party brand. Thus, affective polarization is a self-perpetuating process, providing little room for even elites to move away from the hostility or lies that are sometimes necessary to endorse to signal party loyalty and win elections.

Affective partisanship may not be the necessary result of majoritarian elections, but it may be the necessary result of majoritarian elections with strictly sorted parties. It is difficult to establish whether the same trends would have happened without partisan sorting over the past 60 years, but from the data it seems likely that the two are deeply connected. Furthermore, there is no easy way to undo this partisan sorting, especially given the fact that the incentive structure seems to necessitate it. With this in mind, moving to a more consensus-based system may represent the best option to move towards a better political culture, even if the lack of such a system was not the sole cause of our current situation.

Alternatively, what we are seeing now may simply be the result of a long-term trajectory caused mostly by our use of first-past-the-post elections, hundreds of years in the making. Under this explanation, we may always have been approaching or near the current level of inter-party hatred or simply have gotten here as a result of transitioning from a disparate and largely autonomous union of cultures into one with unified national political parties and 24-hour news. In this case it is difficult to establish a counterfactual, but we can look to other nations experiencing similar changes in media architecture and demographic alignments and see that the long-term trend of those with similar institutions to ours are more extreme than those with more consensus-based systems, especially focusing on polarization.⁷⁴

Our current trajectory may reflect a change in both our political culture and the nongovernmental institutions that buttress an effective democracy all without substantive changes to actual governmental institutions. Regardless of the cause, as shown above, there is ample reason to believe that changes to the institutions of government could have a substantial, positive effect.

While comparative research is vital to establishing counterfactuals, the U.S. is often difficult to compare to other nations. Even among countries with similarly extreme majoritarian systems, few have as strict a two-party system. A change from one electoral system to another in this context is unpredictable and would likely need ample time to fully establish its effects.

Especially of interest would be the effects—if any—of a change in the voting system at a smaller scale. While first-past-the-post elections are still overwhelmingly the norm, certain local jurisdictions or states have begun to experiment with alternative methods, most commonly ranked-choice voting. As these cases become more common or further reaching in their reforms, future research should focus on whether more parties gain representation, how candidates’ rhetoric changes, whether misinformation is less prevalent, and how the average partisan in these areas feels about their opponents. Particularly important will be whether voters perceive these new systems as fairer or their opponents as threats. Given the nationalization of local politics in the U.S., even large changes at the local level may have little effect due to the overwhelming influence of our background national political culture, but even if this is the case, establishing this result firmly would be beneficial to the larger conversation.

Future work should also seek to establish a new baseline for comparison. While it is obvious concerns about electoral fraud and security are widespread, it is not yet clear how durable these perceptions are. Additionally, it is not yet clear to what extent the average American has an opinion on the procedural fairness of the election system beyond the performance of their preferred candidates. Future work should seek not just to investigate the recent misinformation surrounding this issue but establish how important perceptions in this area are generally to Americans.

Currently, a great deal of the research on this topic relies on wide-ranging but unspecific data such as the CSES or the ANES. Such data is useful up to a point and has wide coverage but can be vague for establishing more specific conclusions about feelings towards the government. The questions asked by these surveys often make it difficult to differentiate what respondents are giving their opinion about. For instance, Linde and Ekman found that a question which, on its face, measures support for democratic principles, actually corresponds more to the performance of a specific government.⁷⁵

This lack of specificity is particularly an issue in measures of affective polarization. The most used measurement, the difference in in-versus-out-party affection, inherently represents a great deal of information. As such it is hard to differentiate the effects of differing numbers of parties on affective polarization, as this measure will represent something far different in a two-party system as opposed to a multiparty environment where one's feelings towards other parties might vary wildly. Additionally, research should answer what causes these differences. In a multiparty system out-party hostility may be lower on average because many of the opposition parties might be ideologically like one's own. However, it may also be due to a partisan feeling less threatened by other parties even if they dislike them because the chance of any one of them winning sole power is very low. Disentangling these effects is key to understanding the real effects of reform.

This report does not focus on the actual sources of misinformation, however there is a great deal of work to be done on this subject, especially around manufactured misinformation. Research should aim to answer whether manufactured disinformation is targeted more towards majoritarian systems over consensus-based ones. Due to the importance of who wins in a first-past-the-post system, compared to the high likelihood of coalition and harder to game system of a country that uses a proportional system, it seems likely misinformation campaigns would be less useful in a consensus-based democracy. Research should seek to establish whether these assumptions are sustained under investigation as well as whether there is a significant difference in the amount of targeted misinformation pumped into countries with differing election methods.

To understand misinformation, and ultimately counter it, we must better understand its drivers. Failure to do so has left us with a variety of well-intentioned but ultimately localized responses that may serve as useful band-aids, but will not tackle the root of the misinformation problem. Central to the failure of these reactions to misinformation is an over-emphasis on the supply side of the equation, rather than the factors that ensure politicians, pundits, and voters all demand misinformation. We should worry much more about these sociological drivers than misinformation’s power to fool unsuspecting voters or spread via inherently dubious new technology. Looking at these factors, particularly the power of social identity and affective polarization, leads to a more challenging, but ultimately more productive, understanding that the desire for misinformation is deeply embedded in our political culture and our democracy.

“To understand misinformation, and ultimately counter it, we must better understand its drivers.”

Affective polarization, one of the strongest factors for misinformation sharing, is fed—at least in part—by the incentive structure of oppositional politics. Our natural tendency to construct durable identities for ourselves and the increasing tendency for these identities to overlap with and reinforce our political affiliations means we face more motivation to stereotype, denigrate, lie about, and defend ourselves against our perceived enemies in the opposite party.

While we cannot change many facets of our nature, this report has offered strong evidence that we can change the context in which this process occurs. First-past-the-post voting ensures that a win for me is a loss for you, making every election a no-holds barred battle against bitter enemies. If the rules of our democracy encourage our worst impulses and set us in opposition to one another, our identification with our own party will necessarily become intertwined in a hatred for our opponents. Our rhetoric will become more hostile, our elites more unwilling to compromise, our feelings towards each other more stereotyped and negative, and our information less reliable. While reforming the way we vote is not a panacea to solve all these problems—especially as alternate methods are not immune to developing cultures of political hostility—our current institutions only seem to encourage the political and social conditions that lead us to hate and fear those we disagree with.