Over the past 25 years the partisanship of the American electorate has increased (see e.g., here, here, and here), and concern about how this increased partisanship may impact American democracy is now prevalent (see e.g., here, here, and here).
However, new research by More in Common suggests that Americans are more similar to their political opponents than they realize. These new findings are also consistent with previous research by More in Common which found that very few Americans are staunch ideologies (most, they argued, comprise an ‘exhausted majority’ who hold moderate and heterodox views and are frustrated with how extremes seem to dominate our politics and civic life).
The main findings of their new study, “The Perception Gap,” are also consistent with over three-decades of research. Here, we will briefly present their core findings and connect them to the existing scholarly literature.
Some of the big takeaways from the ‘perception gap’ report Include:
- Democrats and Republicans significantly overestimate how many people on the ‘other side’ hold extreme views. Typically, their estimates are roughly double the actual numbers for a given issue.
- Greater partisanship is associated with holding more exaggerated views of one’s political opponents.
- The Perception Gap is strongest on both “Wings” (America’s more politically partisan groups).
- Consumption of most forms of media, including talk radio, newspapers, social media, and local news, is associated with a wider Perception Gap.
- Education seems to increase, rather than mitigate, the Perception Gap (just as increased education has found to track with increased ideological prejudice). College education results in an especially distorted view of Republicans among liberals in particular.
- The wider people’s Perception Gap, the more likely they are to attribute negative personal qualities (like ‘hateful’ or ‘brainwashed’) to their political opponents.
In other words, these data suggest that Americans know what positions the different political parties typically hold and disagree on (e.g., pro-choice or pro-life; for more government regulation of the economy or for less government regulation of the economy), but they exaggerate the difference between themselves and the constituents of rival parties. The more partisan they are, the more they exaggerate partisan differences (Figure 1).
As can be seen in the table below which documents over two decades of peer-reviewed scholarship, the exaggeration of differences between rival political groups reported by More in Common should not be surprising — nor their finding that greater partisanship increases the exaggeration of these perceived differences — nor their finding that increased information or reflectiveness fails to ameliorate the perception gap.
Overall conclusions about the ‘perception gap’ and its implications follow this brief literature review:
|Publication||Sample Characteristics||Key Findings|
|Judd & Johnson (1981)||54 undergraduate females enrolled at Radcliffe College: 28 feminists and 26 non-feminists.|| “For prevalence estimates of population groups with whom the subject has little direct contact, extreme subjects tend to overestimate the prevalence of extreme positions relative to less extreme subjects. This overestimation is as true for estimates of the disagreeable or anti position as it is for the agreeable or pro position” (p. 33).|
“Subjects with intense affect towards an issue view the world as polarized on that issue” (p. 34).
“Subjects with intense affect towards an issue view positions on that issue as relatively more diagnostic of other personality traits and characteristics. This is true even though they view both pro- and anti-attitudes on that issue as prevalent” (p. 34).
|Robinson et al. (1995)||Study 1a: 27 pro-choice activists and 25 pro-life activists.|
Study 1b: 66 undergraduates enrolled in an introductory psychology course.
Study 2: 59 Stanford undergraduates enrolled in introductory psychology courses or recruited from dormitories; 23 liberals, 20 conservatives, and 16 moderates.
|“… partisans involved in socio-political conflicts tend to overestimate the extremity and ideological congruency of the underlying beliefs and construal of the other side, and often of their own side as well” (p. 414).|
“Another noteworthy, perhaps surprising, feature in our data may be the tendency for those holding the conservative position on the relevant issue to be perceived less accurately than those holding the liberal position, a tendency shared by liberals, by neutrals, and by conservatives themselves” (p. 414).
|Sherman et al. (2003)||Study 1: 78 undergraduates from the dormitories of Stanford University; 58 supported the affirmative action proposal and 14 opposed the proposal. |
Study 2: 53 undergraduates enrolled in an introductory psychology course at Stanford University; 32 supported the affirmative action proposal, 10 were neutral, and 11 opposed the proposal.
Study 3: 58 undergraduates enrolled in introductory psychology courses (32 from Princeton University and 26 from Stanford University); 43 supported the affirmative action proposal and 15 opposed the proposal.
|“Closer examination of Figure 1 reveals that students favoring rejection of the affirmative action proposal were misperceived to a greater extent by allies and opponents alike” (p. 279).|
“Thus, the prediction of a false-polarization effect was clearly confirmed, with both the partisans and nonpartisans assuming the existence of a wide ideological chasm, where little if any existed” (p. 281).
“The fact that in actuality, supporters and rejecters of the affirmative action proposal differed little if at all on ideology and beliefs about other issues, suggests that the assumptions made by participants may have led them astray” (p. 282).
|Chambers & Melnyk (2006)||Study 1: 447 undergraduates enrolled in a psychology course at the University of Florida; 246 had strong pro-choice views and 201 had strong pro-life views.|
Study 2: 480 undergraduates enrolled in an introductory psychology course at the University of Florida; 245 Democrats and 235 Republicans.
|“… the key finding to emerge from the present research was that these (mis)perceptions of disagreement were linked to the specific impressions partisans had of their rivals. In particular, to the extent partisans believed their rivals were opposed to their own core values, they attributed negative traits to members of the rival group, positive traits to members of their own group, saw themselves as less similar to the outgroup, expressed anger toward the outgroup, and exhibited an ingroup bias in favor of their ingroup” (p. 1307).|
|Chambers et al. (2006)||Study 1: 199 undergraduates enrolled in an elementary psychology course at the University of Iowa; 125 were strongly pro-choice and 74 were strongly pro-life. |
Study 2: 88 undergraduates enrolled in an elementary psychology course at the University of Iowa; 29 Democrats, 28 Republicans, and 30 neutral or unaffiliated.
|“Our research suggests that partisans perceive greater disagreement regarding the value issues they see as central to their own position than regarding less central value issues. The partisans in our studies were more alike in their opinions than they knew, and this fact was lost on them because, in their minds, the conflict was not about their adversaries’ central values but their own. Ironically, this led to a situation in which partisans disagreed about what they disagreed about. Each side saw the other as irrationally and stubbornly challenging the very foundation of their personal ideologies, while seeing consensus of opinion about their adversaries’ core values” (p. 43).|
|Graham et al. (2012)||2,212 U.S residents or citizens who visited ProjectImplicit.org: 1,174 liberals, 500 conservatives, and 538 moderates.||“The results also confirm previous studies of misperception by showing that, in general, people overestimate how dramatically liberals and conservatives differ” (p. 8).|
“We presented three competing hypotheses about accuracy:
1) We found some support for the hypothesis that moderates would be most accurate, which they were in the case of the binding foundations. However, and most critically, partisan inaccuracies were not mirror images of each other. On the contrary, liberals and conservatives both tended to exaggerate their binding foundation differences by underestimating the typical liberal and overestimating the typical conservative.
2) We found no support for the hypothesis that liberals would be the most accurate; liberals were the least accurate about conservatives and about liberals. The largest inaccuracies were liberals’ underestimations of conservatives’ Harm and Fairness concerns, and liberals further exaggerated the political differences by overestimating their own such concerns.
3) Finally, we found some support for the hypothesis that conservatives would be the most accurate, which they were in the case of the individualizing foundations. In line with Moral Foundations Theory, liberals dramatically underestimated the Harm and Fairness concerns of conservatives” (p. 9-10).
|Van Boven et al (2012)||Study 1: Nationally representative sample of 1,000 Americans collected in conjunction with the 2008 ANES. |
Study 2: Online survey of 129 U.S. adults recruited from classified advertisements in Atlanta, Denver, Los Angeles, and New York.
Study 3: 28 undergraduates participated for course credit at the University of Colorado, Boulder.
Study 4: 101 undergraduates participated for course credit at the University of Colorado, Boulder.
|“These results demonstrate —for the first time, and with a nationally representative sample — the simultaneous occurrence of polarization projection and simple projection. In the partisan context of a presidential election, people who were more extreme in their own support for one presidential candidate over the other perceived Americans as more polarized compared with people who were moderate in their support for the candidates” (p. 90).|
“Polarization projection and simple projection were also independent of people’s partisan identification. To be sure, people who more strongly identified as a Democrat or Republican perceived more polarization than did people who identified less strongly. But the effects of polarization projection and simple projection were independent of political identification” (p. 90).
“Together, then, these findings indicate that people project onto others perceptions of the processes underlying their own attitudes toward a partisan topic. This process projection occurred even as people perceive others as engaging in less rational, more biased processing than themselves” (p. 91).
“These results conceptually replicate the central findings of Studies 1 and 2, but in a more tightly controlled laboratory setting, with naturally occurring partisan groups and with a novel issue. Participants were confronted with a fictional partisan issue regarding the allocation of resources to naturally occurring groups of resident and nonresident students at a large public university. The more polarized their attitudes, the more they perceived polarization in the distribution of attitudes among other students, independent of simple projection” (p. 94).
“These results indicate that introspection about the processes underlying one’s partisan attitudes increased both polarization projection and simple projection. Introspection increased the relationship between people’s own attitude extremity and their perception of polarized attitudes among others (polarization projection)” (p. 96).
“We found evidence for a psychological phenomenon of polarization projection. People project the extremity of their own partisan attitudes onto others such that those with more extreme attitudes perceive greater polarization than do those with less extreme attitudes” (p. 96).
|Scherer et al. (2015)||Study 1: 204 U.S. residents recruited through Amazon’s Mechanical Turk platform; 72 Democrats, 46 Republicans, 86 Independents/other party. |
Study 2: 219 U.S residents recruited through Amazon’s Mechanical Turk platform
|“… the perceived differences between Republicans and Democrats on these trait measures were significantly larger than the actual differences” (p. 206). |
“… these stereotypes seemed to be driven by an overestimation of the extent to which Republicans possess traits linked to epistemic, existential, and ideological motives. While Democrats seem to have especially exaggerated perceptions of Republicans, there was also evidence that partisans from both ends of the spectrum exhibited greater stereotype exaggeration” (p. 207-208).
|Westfall et al. (2015)||More than 20,000 responses to the ANES, conducted from 1968 to 2008.|| “We found that people perceive greater polarization when partisans are categorized and people estimate the attitudes of the opposing group relative to their own group; when people strongly identify with their partisan group, whether Democrat or Republican; and when people hold relatively extreme partisan attitudes. These last two factors interacted such that people high in issue partisanship — that is, people whose attitudes correspond with their partisan identity—perceived the greatest levels of political attitude polarization” (p. 155).|
“… our results have at least two direct implications for contemporary political discourse. First, nearly all of the effects reported here are symmetrical for Democrats and Republicans. The tendency to overestimate polarization, the association between perceived polarization and political actions, and the factors associated with perceived polarization are true of both Democrats and Republicans” (p. 155).
“Second, among the more striking findings from our analysis is that those who perceive the greatest political attitude polarization in the United States — and, hence, those who most exaggerate political polarization — are those who are themselves most polarized, strongly identifying as party members and holding relatively extreme attitudes that align with their partisan identities” (p. 155).
|Bordalo et al. (2016)||Dataset 1: Graham et al. (2012): 1,174 liberals and 500 conservatives (referred to as the GNH data).|
Dataset 2: More than 20,000 responses to the ANES, conducted from 1964 to 2012.
|“In the full data sets, we treat each (issue, year) pair as an observation, and we cluster standard errors at the issue level. For the GNH data, we have 45 observations: 45 issues measured in the same year. For the ANES data, we have 66 observations: 10 issues, measured in multiple years… Figure IV shows that the believed difference between typical conservative and typical liberal positions is larger than the true difference in mean positions for 109 of the 111 observations. The data for both GNH (squares) and ANES (triangles) lies above the 45-degree line (dashed)” (p. 1785-1787). |
“The systematic and significant exaggeration of mean differences suggests that the benchmark model of accurate beliefs is missing something important. Indeed, this exaggeration reflects the fact that believed means are typically more extreme than true means” (p. 1787).
|Levendusky & Malhotra (2016)||Nationally representative probability sample of 510 U.S. adults, collected by GfK Custom Research.|| “… the perceived divide between Republicans and Democrats on every issue is larger than the actual divide between Republicans and Democrats” (p. 384).|
“Figure 1 shows that even though people view members of their own party as more extreme than they actually are, they view opposing partisans as even more extreme (see also table 1)” (p. 386).
“Hence, false polarization is caused by people perceiving both their own party and the opposing party to be more extreme than they are in reality, but their perceptions of the extremity of opposing partisans are even more skewed” (p. 388).
|Lammers et al. (2017)||Study 1: 114 Americans recruited through Amazon’s Mechanical Turk.|
Study 2: 202 Americans recruited through Amazon’s Mechanical Turk.
Study 3: 300 Americans recruited through Amazon’s Mechanical Turk.
Study 4: 300 Americans recruited through Amazon’s Mechanical Turk.
Study 5: 300 Americans recruited through Amazon’s Mechanical Turk.
Study 6: 300 Americans recruited through Amazon’s Mechanical Turk.
Study 7: 483 Americans recruited through Amazon’s Mechanical Turk.
Study 8: 302 Americans recruited through Amazon’s Mechanical Turk.
Meta-analysis: Performed on all 8 studies and on 2 additional studies that did not show the expected effect.
|“People on both political extremes categorize stimuli in the political domain more strongly than do moderates. They are more likely to cluster similar political stimuli closer together and form tighter, more homogenous categories. This applies to a wide range of political stimuli, such as politicians (Study 1), groups associated with different ideologies (Study 2), or newspapers (Study 3), and also applies to the inferences that people make about politics — for example, how people vote across the country (Study 4) and whether politically different people share personal tastes (Study 5) or social relationships (Study 6-8)” (p. 618-619).|
“The current work extends beyond existing work by its basic nature. Earlier research shows that people on the political extremes exaggerate differences across the political divide, see greater polarization, in their opponents’ attitudes, and are more dogmatic. The current findings fit with that literature but go beyond it by showing that people with strong political opinions not only have more extreme views but even represent the political domain differently. They categorize the same stimuli differently and perceive more homogenous and separate categories than do moderates and neutrals” (p. 619).
“When differences between political groups or stimuli are large and real, drawing sharper political clusters and categories can increase the efficient use of cognitive resources, allowing people on the political extremes to draw inferences more quickly than the politically moderate. On the other hand, it may lead people to exaggerate differences between categories and thus introduce oversimplified and erroneous thinking about a complex and multifaceted word. This explains why liberals and conservatives in the United States find it so difficult to connect. People on the political extremes may not only miss opportunities to connect across the partisan divide because of negative emotions or lack of motivation but also because in their perception of political reality there simply is no middle ground” (p. 619).
|Van Boven et al. (2018)|| Sample 1: 1,056 Americans recruited through Qualtrics’ national panel.|
Sample 2: 1,065 Americans recruited through ROI Rocket’s proprietary “general population” market research panel.
|“As much as people are personally swayed by partisanship, they expect partisan influence to be even stronger among other Democrats and Republicans. People generally exaggerate partisan opposition, thereby overestimating political polarization between Democrats and Republicans. Whereas people see their own stances as grounded in careful, dispassionate, and nonideological analysis, they see others’ stances as blatantly partisan, often reflecting careless, ideological thinking. In other words, people exaggerate how much partisans reactively devalue the opposing side’s ideas” (p. 500).|
“Respondents slightly exaggerated how much their fellow Democrats and Republicans would support policies from their own political party … To a much larger degree, respondents exaggerated reactive devaluation by partisans on the other side. Respondents substantially underestimated Republicans’ support for Democratic policies … and Democrats’ support for Republican policies … Thus, even though Democrats and Republicans were themselves swayed by partisan framing, they exaggerated how much the degree to which partisan framing would influence others’ policy support” (p. 500).
“People support the same policies when proposed by their own party more strongly than when proposed by the other party, and people exaggerate how much other Democrats and Republicans are swayed by partisanship, which creates potent, but false, norms of partisan opposition” (p. 503).
The findings reported by More in Common should be treated as an additional data point for the table above. All studies have flaws and all of the studies listed here can be criticized to some degree. Some have small sample sizes and/or non-representative samples. Others, that made an effort to obtain representative probability samples, can be critiqued on how the questions are worded.
Thus, on the one hand it is not surprising that More in Common’s conclusions have already been criticized. Critique of methodology and statistical analysis is always welcome, they are one of the primary driving forces of scientific investigation. On the other hand, however, the findings reported by More in Common are remarkably similar to each of the studies documented above. Furthermore, the studies documented in Table 1 run the gamut from small unrepresentative samples, to larger unrepresentative samples, to nationally representative probability samples, and finally to cross-sectional analyses of the American National Election Study. That so many different samples yield similar results, lends a good deal of support to the conclusions that More in Common reached.
The political Perception Gap appears to be very real. Ignoring or dismissing this fact is not going to reduce it. In fact, as suggested by Chambers & Melnyk (2006) and Chambers et al. (2006), it may exacerbate it. Thus, a more salient question may be, “what can be done about it?” given that many of the most intuitively obvious correctives — education, the media (including from flagship publications like the New York Times, Washington Post et al.) and political engagement — all seem to make the problem worse.
Social media may also contribute to the perception gap by pressuring people to publicly conform to the expected position of their tribe despite the reality that they may hold much more nuanced, even heretical, views (which are often revealed in more private settings); the More in Common study finds that increased sharing of political content on social media correlates with a higher partisan perception gap.
It may be possible to ‘fix’ higher ed and media institutions so that they become part of the solution instead. This is, for instance, what Heterodox Academy hopes to do with respect to institutions of higher learning (and organizations like Solutions Journalism aspire to do for the media).
However, a number of organizations and movements nationwide have also emerged to help people with different commitments, and from different walks of life, engage with one another directly (e.g. The Village Square, Better Angels, Bridge USA, Free Intelligent Conversations — more here). Although these latter efforts have often proven challenging to scale, they offer a means for citizens to sidestep the institutions currently mediating (and often undermining) our engagement with one-another.
It seems plausible that a ‘both/ and’ approach may be most effective here, although more applied research on how to effectively mitigate partisan perception gaps would be helpful.
Citation: Daniel Yudkin et al. (2019). “The Perception Gap: How False Impressions are Pulling Americans Apart.” More in Common, June.
Sean Stevens is Heterodox Academy’s Research Director. He has a PhD in social psychology from Rutgers University.
As an organization that prizes pluralism and disagreement — with more than 2,500 members holding diverse views on most issues — Heterodox Academy almost never takes positions as an organization on current events and controversies. Opinions expressed here are those of the author(s). Publication does not imply endorsement by Heterodox Academy or any of its members. We encourage readers to follow us on Facebook, Twitter and LinkedIn — and to join in the conversation on those forums — to weigh in on this or other posts.