Abstract

Few topics in public opinion research have attracted as much attention in recent years as partisan polarization in the American mass public. Yet, there has been considerably less investigation into whether people perceive the electorate to be polarized and the patterns of these perceptions. Building on work in social psychology, we argue that Americans perceive more polarization with respect to policy issues than actually exists, a phenomenon known as false polarization. Data from a nationally representative probability sample and a novel estimation strategy to make inferences about false polarization show that people significantly misperceive the public to be more divided along partisan lines than it is in reality. Also, people’s misperceptions of opposing partisans are larger than those about their own party. We discuss the implications of these empirical patterns for American electoral politics.

Scholarly debate over mass partisan polarization 1 in the American public has been extensive in recent years (see Fiorina and Abrams [2008] for a review). An important, related topic that has received less attention is whether Americans perceive themselves to be polarized along partisan lines. Social psychologists have documented a phenomenon known as “false polarization” (e.g., Pronin, Puccio, and Ross 2002 ; Sherman, Nelson, and Ross 2003 ; Chambers, Baron, and Inman 2006 ), or people’s tendency to overestimate the degree of polarization between groups. False polarization is the difference between two quantities: (1) the distance between the perceived positions of Group A and Group B; and (2) the distance between the actual positions of Group A and Group B. False polarization occurs because of the underlying psychology of classification. When people view themselves as part of groups and believe these identities to be salient, perceived differences between groups are accentuated ( Tajfel and Wilkes 1963 ).

This research note documents the extent and characteristics of false polarization among the American mass public, leveraging for the first time a nationally representative probability sample. Additionally, we introduce a new estimation strategy for making statistical inferences about false polarization. We find evidence of false polarization in the American public, show that people view opposing partisans as more extreme than fellow co-partisans, and discuss the implications of these empirical patterns for American electoral politics.

Literature Review

Many studies have documented false polarization along party lines. For instance, Van Boven, Judd, and Sherman (2012) showed in convenience samples of undergraduates and subjects recruited via online classified ads that Republicans and Democrats perceive a far greater divide between the parties in the mass public than actually exists. 2 However, it is important to assess whether these trends exist among representative, probability samples of the American public. Studies have shown that opt-in convenience samples differ substantially from representative samples. In particular, individuals from opt-in samples are much more politically interested and engaged ( Malhotra and Krosnick 2007 ; Yeager et al. 2011 ). Because political expertise is positively correlated with extremity ( Federico 2004 ), people who are more passionate and knowledgeable about politics—such as opt-in survey respondents—may project their own, more extreme opinions onto others ( Goel, Mason, and Watts 2010 ). Highly interested individuals are also more likely to consume partisan media ( Stroud 2011 ) and therefore may be selectively exposed to extreme voices. Thus, estimates of false polarization from convenience samples may be biased upward. Probability sampling is also important for estimating false polarization because one of the components of the measure is the actual distance between the mass parties. If this benchmark is not accurately estimated, then false polarization cannot be measured with any precision.

There have been attempts to use the high-quality American National Election Studies (ANES) to document false polarization ( Brady and Sniderman 1985 ; Westfall et al. 2012 ). While the ANES asks respondents to place themselves on ideological scales, it does not ask people to place their fellow citizens on these same scales. It only asks people to place “the Republican Party” or “the Democratic Party.” However, when people are reporting their views on the “party,” they may be thinking about political elites as opposed to their fellow citizens in the mass public. As a result, estimates of false polarization from the ANES may simply be picking up the “disconnect” between elites and masses ( Fiorina and Levendusky 2006 ). Westfall et al. (2012) present evidence from Amazon’s Mechanical Turk that when people are evaluating the “parties,” they may be thinking about parties in the mass public, but our goal here is to bring direct evidence from nationally representative data to the question.

Data and Survey Items

We analyze a high-quality, nationally representative probability sample from GfK Custom Research (formerly Knowledge Networks) conducted between November 29 and December 12, 2012. GfK uses random-digit dialing and address-based sampling methods to recruit panelists to take surveys such as ours; the resulting data are therefore a random sample of the US adult population. The target population was US adults. All interviews were conducted in English. A total of 510 panelists completed the survey, generating a completion rate of 64.5 percent (AAPOR cumulative response rate [CUMRR1] of 6.3 percent).

Our setup and design are similar to earlier studies of false polarization, but we used a new set of issues (taxes, immigration, free trade, and public financing 3 ), and estimated our quantities of interest using nationally representative data and a novel empirical strategy.

We present below an example (in the context of capital gains taxes) that illustrates the three survey items (in addition to party identification) needed to measure false polarization: (1) a respondent’s self-placement on the issue scale; (2) placement of the typical Democratic voter on that scale; and (3) placement of the typical Republican voter on that scale. Respondents were asked:

The tax rates on the profits people make from selling stocks and bonds, called capital gains taxes, are currently lower than the income tax rates many people pay. Do you think that capital gains tax rates should be increased, decreased, or kept about the same?

Respondents were provided a seven-point scale, presented horizontally, similar to the ANES’s questions on ideological placement. The scale was fully labeled; the response options were “increased a lot,” “increased somewhat,” “increased a little,” “kept the same,” “decreased a little,” “decreased somewhat,” and “decreased a lot.” All items were recoded to lie between 0 and 1, with the most liberal response coded as 0 and the most conservative response coded as 1. In addition to placing themselves on the scale, they also placed the typical Republican and Democratic voter:

What do you think the TYPICAL [DEMOCRATIC/REPUBLICAN] VOTER would want to happen to capital gains tax rates?

Respondents were provided the same seven-point scale as the self-placement item. All four issues were asked on seven-point scales. The appendix provides the full set of question wordings and response options.

We measure partisanship using a standard branching design, where subjects are first asked their basic orientation, then the strength of that affiliation (or, if an Independent, whether they lean toward either party; see appendix for exact wording). Throughout the analysis below, we include leaners as partisans (per Keith et al. 1992 ). Excluding them does not change our substantive results (see online appendix A). 4

Two features of our survey might have influenced the results. First, the survey was administered shortly after the 2012 elections, when respondents may have felt lingering animosity due to a closely contested election. Second, because respondents first report their own positions on the issue, then report their perceptions of other partisans, their self-placements might serve as an anchor and influence their placement of others ( Conover and Feldman 1982 ; Brady and Sniderman 1985 ). We adopted this measurement strategy to be broadly consistent with previous studies. In a second survey in August 2014 using a convenience sample to investigate both concerns (see online appendix B), we find that while these measurement issues do affect the size of the false polarization effect, the results are not simply an artifact of survey timing or anchoring. 5

Complete descriptive statistics for these survey items are presented in table 1 for the four issues and for an index averaging responses to the four issues together. Demographic characteristics of the sample can be found in online appendix C.

Table 1.

Descriptive Statistics of Real and Perceived Issue Positions

 Actual positions  Perceived positions
(all respondents)  
Perceived positions
(Democratic respondents)  
Perceived positions
(Republican respondents)  
Ds Rs Ds Rs Ds Rs Ds Rs 
Average  .40
( n = 263)  
.58
( n = 225)  
.31
( n = 501)  
.69
( n = 501)  
.36
( n = 262)  
.72
( n = 262)  
.25
( n = 225)  
.66
( n = 225)  
Taxes  .35
( n = 259)  
.50
( n = 224)  
.29
( n = 496)  
.65
( n = 492)  
.33
( n = 261)  
.68
( n = 257)  
.24
( n = 224)  
.61
( n = 223)  
Immigration  .35
( n = 261)  
.58
( n = 223)  
.30
( n = 499)  
.74
( n = 497)  
.36
( n = 260)  
.78
( n = 259)  
.21
( n = 225)  
.70
( n = 225)  
Trade  .49
( n = 126)  
.58
( n = 120)  
.39
( n = 247)  
.68
( n = 247)  
.41
( n = 125)  
.71
( n = 125)  
.36
( n = 119)  
.66
( n = 118)  
Public financing  .50
( n = 134)  
.73
( n = 105)  
.31
( n = 247)  
.66
( n = 247)  
.35
( n = 133)  
.66
( n = 132)  
.24
( n = 105)  
.67
( n = 105)  
 Actual positions  Perceived positions
(all respondents)  
Perceived positions
(Democratic respondents)  
Perceived positions
(Republican respondents)  
Ds Rs Ds Rs Ds Rs Ds Rs 
Average  .40
( n = 263)  
.58
( n = 225)  
.31
( n = 501)  
.69
( n = 501)  
.36
( n = 262)  
.72
( n = 262)  
.25
( n = 225)  
.66
( n = 225)  
Taxes  .35
( n = 259)  
.50
( n = 224)  
.29
( n = 496)  
.65
( n = 492)  
.33
( n = 261)  
.68
( n = 257)  
.24
( n = 224)  
.61
( n = 223)  
Immigration  .35
( n = 261)  
.58
( n = 223)  
.30
( n = 499)  
.74
( n = 497)  
.36
( n = 260)  
.78
( n = 259)  
.21
( n = 225)  
.70
( n = 225)  
Trade  .49
( n = 126)  
.58
( n = 120)  
.39
( n = 247)  
.68
( n = 247)  
.41
( n = 125)  
.71
( n = 125)  
.36
( n = 119)  
.66
( n = 118)  
Public financing  .50
( n = 134)  
.73
( n = 105)  
.31
( n = 247)  
.66
( n = 247)  
.35
( n = 133)  
.66
( n = 132)  
.24
( n = 105)  
.67
( n = 105)  

N ote .—The columns labeled “Ds” denote the actual or perceived positions of Democratic respondents; those labeled “Rs” denote the equivalent for Republican respondents.

Statistical Methods

Analyzing the data is non-obvious, given that we are comparing a quantity measured for an individual survey respondent (perceived polarization) and a quantity measured across groups of respondents (actual polarization). Indeed, no existing study on false polarization has been able to assess whether the effects are statistically significant, requiring us to develop a novel method.

To estimate the false polarization effect, we first transform the data such that each respondent contributes two observations: (1) their own position on the seven-point scale; and (2) the difference between where they place Democratic and Republican partisans on the seven-point scale. We then estimate ordinary least squares (OLS) regressions of the form

 
(1)
Yi=βOi+γ(Oi×Ri)+ηPi+εi,

where i indexes respondents, Yi represents a survey response provided by the respondent, Oi represents a dummy variable indicating whether the response is the respondent’s own position, Ri is a dummy variable indicating whether the respondent is a Republican (Democrats are the baseline category), Pi is a dummy variable indicating whether the response is the respondent’s perceived position of partisans, 6 and εi is normally distributed stochastic error. Standard errors are clustered by respondent, given that respondents contribute more than one observation in the analysis.

The parameters of equation (1) can be interpreted as follows. First, note that there is no constant term in the regression. Therefore, β represents the mean position of Democrats in the sample, and β + γ represents the mean position of Republicans in the sample. The level of actual polarization is therefore simply represented by the difference, γ . The average perceived polarization in the full sample (combining Republicans and Democrats) is represented by η . Hence, the linear combination ηγ represents the false polarization effect, and a Wald test can determine if this quantity is significantly different from zero. We can also employ a nonlinear Wald test to evaluate whether η / γ = 1 (i.e., is the ratio of perceived polarization to actual polarization greater than 1?).

To assess whether there is partisan asymmetry in perceptions (i.e., if people think that opposing partisans are more extreme than co-partisans), we estimate the following OLS regression:

 
(2)
Yi=βOi+γ(Oi×Ri)+ηPRi+δPDi+φ(PRi×Ri)+λ(PDi×Ri)+εi,

where the variables are defined as above except that we separately enter in the respondent’s perceived position of Republicans ( PRi ) and her perceived position of Democrats ( PDi ).

Equation (2) is a bit more complex. As before, β represents the actual mean position of Democratic respondents and β + γ represents the actual mean position of Republican respondents. Democrats’ misperceptions of fellow Democrats are therefore represented by δ − β , and Democrats’ misperceptions of Republicans are represented by η − ( β + γ ). The partisan asymmetry in misperceptions among Democratic respondents is consequently the difference between these two linear combinations: η − γ − δ . A Wald test also determines if this estimand is significantly different from zero. The algebra for Republicans is more involved but still straightforward. Republicans’ misperceptions of fellow Republicans are represented by ( η + φ ) ( β + γ ), and Republicans’ misperceptions of Democrats are represented by ( δ + λ ) − β . Therefore, the partisan asymmetry in misperceptions among Republican respondents is the difference between these two quantities: δ − λ − η − φ − γ . Note that we construct these differences such that the sign of the bias is always positive (e.g., subtracting Democrats’ likely lower misperception of co-partisans’ level of extremity from their higher misperceptions of Republicans’ level of extremity).

Results

AMERICANS PERCEIVE GREATER POLARIZATION THAN ACTUALLY EXISTS

As shown in figure 1 (see also table 1 ), the perceived divide between Republicans and Democrats on every issue is larger than the actual divide between Republicans and Democrats. Averaging across the four issues (see the top row of the figure), the actual distance between the mass parties (represented by the heavy black line) on the normalized 0–1 scale is 0.18. Republicans place themselves at 0.58 (somewhat to the right of center), and Democrats place themselves at 0.40 (somewhat to the left of center). However, respondents perceived people of different parties to be 0.38 units apart (see the thin black line): Republicans are perceived to be at 0.69, and Democrats are perceived to be at 0.31. Hence, the mass parties are perceived to be about 0.20 units (or 20 percent) further apart than they were in reality.

Figure 1.

Perceived vs. Actual Polarization.

Figure 1.

Perceived vs. Actual Polarization.

Table 2 shows that the false polarization effect is statistically significant ( t = 12.8, p < 0.01)—the gap between perceptions and reality can be distinguished from zero. As shown in the remainder of figure 1 and tables 1 and 2 , this pattern is also true for each individual issue. The last column of table 2 presents the ratio between perceived and actual polarization. For each test statistic, we reject the null hypothesis that the ratio is equal to 1 at p < .05. Although people are relatively moderate on average, they see others as more extreme than they are in reality. 7

Table 2.

Americans Perceive More Polarization Than There Actually Exists

   False polarization 
 Perceived polarization Actual polarization Perceived− Actual Perceived/ Actual 
Average  .380
(.015)  
.175
(.016)  
.205
(.016)  
2.17
(.18)  
Taxes  .361
(.020)  
.152
(.021)  
.209
(.021)  
2.37
(.29)  
Immigration  .454
(.018)  
.227
(.027)  
.228
(.026)  
2.00
(.22)  
Trade  .302
(.028)  
.084
(.032)  
.217
(.033)  
3.58
(1.25)  
Public financing  .362
(.032)  
.231
(.041)  
.131
(.038)  
1.57
(.24)  
   False polarization 
 Perceived polarization Actual polarization Perceived− Actual Perceived/ Actual 
Average  .380
(.015)  
.175
(.016)  
.205
(.016)  
2.17
(.18)  
Taxes  .361
(.020)  
.152
(.021)  
.209
(.021)  
2.37
(.29)  
Immigration  .454
(.018)  
.227
(.027)  
.228
(.026)  
2.00
(.22)  
Trade  .302
(.028)  
.084
(.032)  
.217
(.033)  
3.58
(1.25)  
Public financing  .362
(.032)  
.231
(.041)  
.131
(.038)  
1.57
(.24)  

N ote .—Standard errors in parentheses.

PEOPLE ARE MORE LIKELY TO PERCEIVE OPPOSING PARTISANS AS MORE EXTREME THAN CO-PARTISANS

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 ). As shown in the top row of the figure, averaging across all four issues, Republicans placed fellow Republicans at 0.66 (see the red-colored “R” 8 below the line), further to the right than Republicans’ actual position at 0.58. However, Democrats placed Republicans even further to the right, at 0.72 (see the blue-colored “R” above the line). Similarly, Democrats placed fellow Democrats at 0.36 (more extreme than their actual position of 0.40; see the blue-colored “D” above the line), but Republicans placed Democrats at an even more extreme point (0.25; see the red-colored “D” below the line). The level of partisan asymmetry among Republican respondents is therefore 0.07 units (they misperceive Democrats as 0.15 units more extreme than they truly are, but Republicans as only 0.08 units more extreme). As shown in table 3 , this difference can be statistically distinguished from zero ( t = 3.2, p = .002). Similarly, the level of partisan asymmetry among Democratic respondents is 0.094 units (they misperceive Republicans as 0.138 units more extreme than they truly are, but Democrats as only 0.044 units more extreme). Again, this quantity is statistically significant ( t = 4.9, p < .001). 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. 9

Table 3.

Partisan Bias in Misperceptions

 Democratic respondents Republican respondents 
Misperception of Republicans Misperception of Democrats Difference Misperception of Democrats Misperception of Republicans Difference 
Average  .138
(.017)  
.044
(.010)  
.094
(.019)  
.149
(.018)  
.083
(.013)  
.067
(.021)  
Taxes  .186
(.023)  
.012
(.014)  
.174
(.026)  
.107
(.024)  
.110
(.016)  
−.003
(.026)  
Immigration  .200
(.025)  
−.010
(.016)  
.210
(.030)  
.142
(.025)  
.123
(.020)  
.019
(.031)  
Trade  .130
(.034)  
.087
(.024)  
.044
(.041)  
.131
(.032)  
.085
(.024)  
.046
(.035)  
Public financing  −.078
(.041)  
.153
(.026)  
−.231
(.049)  
.262
(.036)  
−.060
(.030)  
.323
(.046)  
 Democratic respondents Republican respondents 
Misperception of Republicans Misperception of Democrats Difference Misperception of Democrats Misperception of Republicans Difference 
Average  .138
(.017)  
.044
(.010)  
.094
(.019)  
.149
(.018)  
.083
(.013)  
.067
(.021)  
Taxes  .186
(.023)  
.012
(.014)  
.174
(.026)  
.107
(.024)  
.110
(.016)  
−.003
(.026)  
Immigration  .200
(.025)  
−.010
(.016)  
.210
(.030)  
.142
(.025)  
.123
(.020)  
.019
(.031)  
Trade  .130
(.034)  
.087
(.024)  
.044
(.041)  
.131
(.032)  
.085
(.024)  
.046
(.035)  
Public financing  −.078
(.041)  
.153
(.026)  
−.231
(.049)  
.262
(.036)  
−.060
(.030)  
.323
(.046)  

N ote .—Standard errors in parentheses.

Conclusions

This research note has explored whether Americans misperceive the extent of partisan polarization in America, as well as features of those misperceptions. But why do Americans believe that they are further apart than they actually are? Several hypotheses can be tested in future research. For instance, perhaps people are predominantly exposed to stereotypes of opposing partisans through the mass media and therefore employ the availability heuristic in constructing their judgments. Alternatively, the effect may be due to projection, with people placing people from an outgroup far away from them ( Brody and Page 1972 ). But why would they place fellow partisans away from them as well? This could be because moderation is a well-respected trait and people like to see themselves as middle-of-the-road and therefore make advantageous comparisons with others ( Dennis 1988 ; Klar and Krupnikov 2013 ). Alternatively, the minority of extreme voices might be most salient ( Prentice and Miller 1993 ; Noelle-Neumann 1974 ).

Future research could also explore how these patterns vary across different types of issues. 10 We chose moderately salient issues here, but perhaps we would observe different patterns for issues that were more or less salient, or where elite partisan messages (e.g., party platforms, campaign speeches), especially as transmitted through the mass media, were more or less clear. How the dynamics of the issue, as well as the broader media and information environment, shape the level of perceived (and hence false) polarization is an important area for future research.

What are the consequences of these misperceptions? It is possible that misperceiving the other side (and one’s own side) as more extreme may detach people from the political process, decreasing their participation as they feel they are unrepresentative moderates in the extremist milieu of American politics. Because they perceive opposing partisans as more extreme, they may denigrate the other party and lower their affect toward them ( Iyengar, Sood, and Lelkes 2012 ). It is also possible that these misperceptions can affect people’s own attitudes, although the direction is unclear ( Ahler 2014 ; Levendusky and Malhotra forthcoming). Armed with strong evidence of false polarization and its patterns, we look forward to future research exploring the downstream consequences of these misperceptions.

Supplementary Data

Supplementary data are freely available online at http://poq.oxfordjournals.org/ .

Appendix: Question Wording, GfK Data

Party Identification: “Generally speaking, I think of myself as a:” (response options: “Democrat,” “Republican,” “Independent”). If respondents said they were Democrats or Republicans, they were asked, “Would you call yourself a strong [Democrat/Republican] or a not very strong [Democrat/Republican]?” Respondents who indicated that they were Independents were asked, “Do you think of yourself as closer to the Democratic Party or closer to the Republican Party?”

Capital Gains Taxes: “The tax rates on the profits people make from selling stocks and bonds, called capital gains taxes, are currently lower than the income tax rates many people pay. Do you think that capital gains tax rates should be increased, decreased, or kept about the same?” (response options: “Increased a lot,” “Increased somewhat,” “Increased a little,” “Kept the same,” “Decreased a little,” “Decreased somewhat,” “Decreased a lot”)

Typical Democrat/Republican: “What do you think the TYPICAL DEMOCRATIC VOTER would want to happen to capital gains tax rates?” The response options are the same as the respondent’s self-placement above, and the item is repeated for the typical Republican voter. Respondents are randomly assigned to answer either the “typical Democrat” or the “typical Republican” prompt first. Subjects respond to the parallel items for the immigration, trade, and public financing items below.

Immigration: “There is some debate about whether or not undocumented immigrants who were brought to this country illegally as children should be deported. Which of the following positions on the scale below best represents your position on this issue?” (response options: “Very strongly oppose deportation,” “Somewhat strongly oppose deportation,” “Not so strongly oppose deportation,” “Neither support nor oppose deportation,” “Not so strongly support deportation,” “Somewhat strongly support deportation,” “Very strongly support deportation”) All respondents answered both the immigration and capital gains taxes items; the order of these items was randomized.

Free Trade: “The United States is currently considering signing additional free trade agreements with Central American, South American, and Asian countries. The Democratic Party wants to make it more difficult for the United States to enter into such agreements. The Republican Party wants to make it easier to do so. What do you think? Do you support or oppose the United States signing more free trade agreements with Central American, South American, and Asian countries?” (response options: “Very strongly oppose trade,” “Somewhat strongly oppose trade,” “Not so strongly oppose trade,” “Neither support nor oppose trade,” “Not so strongly support trade,” “Somewhat strongly support trade,” “Very strongly support trade”)

Public Election Financing: “Public financing of elections is when the government pays for the cost of campaigning for various offices, rather than the campaigns relying on donations from the general public, corporations, or unions. Democrats typically support public financing plans, while Republicans have wanted to eliminate them. What do you think? Do you support or oppose the government paying for the public financing of elections?” (response options: “Very strongly oppose public financing,” “Somewhat strongly oppose public financing,” “Not so strongly oppose public financing,” “Neither support nor oppose public financing,” “Not so strongly support public financing,” “Somewhat strongly support public financing,” “Very strongly support public financing”) Subjects were randomly assigned to answer either the trade or public election financing items.

References

Ahler
Douglas J
.
2014
.
“Self-Fulfilling Misperceptions of Public Polarization.”
Journal of Politics
 
76
:
607
20
.
Brady
Henry
Sniderman
Paul
.
1985
.
“Attitude Attribution: A Group Basis for Political Reasoning.”
American Political Science Review
 
79
:
1061
1078
.
Brody
Richard
Page
Benjamin
.
1972
.
“Comment: The Assessment of Policy Voting.”
American Political Science Review
 
66
:
450
58
.
Chambers
John
Baron
Robert
Inman
Mary
.
2006
.
“Misperceptions in Intergroup Conflict: Disagreeing about What We Disagree About.”
Psychological Science
 
17
:
38
45
.
Conover
Pamela Johnston
Stanley
Feldman
.
1982
.
“Projection and the Perception of Candidates’ Issue Positions.”
Western Political Quarterly
 
35
:
228
44
.
Dennis
Jack
.
1988
.
“Political Independence in American, Part I: On Being an Independent Partisan Supporter.”
British Journal of Political Science
 
18
:
77
109
.
Federico
Christopher
.
2004
.
“Predicting Attitude Extremity: The Interactive Effects of Schema Development and the Need to Evaluate and Their Mediation by Evaluative Integration.”
Personality and Social Psychology Bulletin
 
30
:
1281
1294
.
Fiorina
Morris
Abrams
Samuel
.
2008
.
“Political Polarization in the Mass Public.”
Annual Review of Political Science
 
11
:
563
88
.
Fiorina
Morris
Matthew
Levendusky
.
2006
.
“Disconnected: The Political Class versus the People.”
In
Red and Blue Nation? Characteristics, Causes, and Consequences of America’s Polarized Politics
  , edited by
David
Brady
and
Pietro
Nivola
,
49
71
.
Washington, DC
:
Brookings Institution Press and the Hoover Institution Press
.
Goel
Sharad
Mason
Winter
Watts
Duncan
.
2010
.
“Real and Perceived Attitude Agreement in Social Networks.”
Journal of Personality and Social Psychology
 
99
:
611
21
.
Iyengar
Shanto
Sood
Gaurav
Lelkes
Yphtach
.
2012
.
“Affect, Not Ideology: A Social Identity Perspective on Polarization.”
Public Opinion Quarterly
 
76
:
405
31
.
Keith
Bruce E.
Magleby
David B.
Nelson
Candice J.
Orr
Elizabeth
Westlye
Marc C.
Wolfinger
Raymond E.
.
1992
.
The Myth of the Independent Voter
  .
Berkeley
:
University of California Press
.
Klar
Samara
Krupnikov
Yanna
.
2013
.
“The Social Consequences of Partisan Disagreement.”
 
Paper presented at the Annual Meeting of the American Political Science Association
,
Chicago, IL, USA
.
Levendusky
Matthew
Malhotra
Neil
.
Forthcoming. “Does Media Coverage of Partisan Polarization Affect Political Attitudes?”
Political Communication
  .
Malhotra
Neil
Krosnick
Jon
.
2007
.
“The Effect of Survey Mode and Sampling on Inferences about Political Attitudes and Behavior: Comparing the 2000 and 2004 ANES to Internet Surveys with Nonprobability Samples.”
Political Analysis
 
15
:
286
323
.
Noelle-Neumann
Elisabeth
.
1974
.
“The Spiral of Silence: A Theory of Public Opinion.”
Journal of Communication
 
24
:
43
51
.
Prentice
Deborah A.
Miller
Dale T.
.
1993
.
“Pluralistic Ignorance and Alcohol Use on Campus: Some Consequences of Misperceiving the Social Norm.”
Journal of Personality and Social Psychology
 
64
:
243
56
.
Pronin
Emily
Puccio
Carolyn
Ross
Lee
.
2002
.
“Understanding Misunderstanding: Social Psychological Perspectives.”
In
Heuristics and Biases: The Psychology of Intuitive Judgemen t
  , edited by
Thomas
Gilovich
Dale
Griffin
Daniel
Kahneman
,
636
65
.
New York
:
Cambridge University Press
.
Sherman
David
Nelson
Leif
Ross
Lee
.
2003
.
“Naive Realism and Affirmative Action: Adversaries Are More Similar Than They Think.”
Basic and Applied Social Psychology
 
25
:
275
89
.
Stroud
Natalie J
.
2011
.
Niche News: The Politics of News Choice
  .
New York
:
Oxford University Press
.
Tajfel
Henri
Wilkes
A. L.
.
1963
.
“Classification and Quantitative Judgment.”
British Journal of Psychology
 
54
:
101
14
.
Van Boven
Leaf
Charles
Judd
Sherman
David
.
2012
.
“Political Polarization Projection: Social Projections of Partisan Attitude Extremity and Attitudinal Processes.”
Journal of Personality and Social Psychology
 
103
:
84
100
.
Westfall
Jacob
John
Chambers
Charles
Judd
van Boven
Leaf
.
2012
.
“The Exaggeration of Political Polarization.”
 
Manuscript, University of Colorado–Boulder
.
Yeager
David
Jon
Krosnick
LinChiat
Chang
Harold
Javitz
Matthew
Levendusky
Alberto
Simpser
Wang
Rui
.
2011
.
“Comparing the Accuracy of RDD Telephone Surveys and Internet Surveys Conducted Probability and Non-Probability Samples.”
Public Opinion Quarterly
 
75
:
709
47
.
1
There is a large literature on what “polarization” actually means ( Fiorina and Abrams 2008 ). Here, we mean “partisan polarization,” or the difference between the issue positions of Democrats and Republicans ( Fiorina and Levendusky 2006 ).
2
Study 1 in Van Boven, Judd, and Sherman (2012) uses a nationally representative sample, but asks about the opinions of Americans in general, not differentiated by party. This study therefore cannot examine perceived partisan polarization, which is our focus here. Their other studies, which allow an assessment of perceived partisan polarization, use convenience samples.
3
All respondents answered the items related to taxes and immigration. Respondents were randomly assigned to answer questions about either the free trade or the public financing issue. These latter two items were included as a party cue explaining the national party’s position on these issues, which allowed us to explore whether the cue would affect the observed extent of false polarization. Because the party cue did not change the basic pattern of false polarization presented in figure 1 and table 2 , we do not discuss it further here.
4
The small percentage of respondents that did not respond to the leaning question (3.5 percent) were excluded from analysis.
5
For instance, we re-estimated the false polarization effect using different methodological approaches: (1) temporally separating respondents’ self-placements and their placement of other partisans using a distractor task; (2) asking respondents to place other partisans before placing themselves; and (3) randomly assigning respondents to either provide self-placements or placements of other partisans. In all cases, the false polarization effect was smaller than that reported here but was not eliminated.
6
Note that an equivalent model could organize the data such that respondents each provided three observations (their own position, their perceived position of Republicans, and their perceived positions of Democrats), and included dummies for each of the two parties separately. This model produces the same parameter estimates. We leverage this model when assessing the degree of partisan asymmetry.
7
Online appendix D presents some descriptive statistics assessing which respondents are most likely to misperceive. Results show that false polarization is increasing in education and issue extremity, and white respondents are more likely to misperceive compared to non-white respondents.
8
In the print version of figure 1 , the red characters representing Republicans’ perceptions are given below the line, and the blue characters representing Democrats positions are given above the line. For a full-color version of the figure, please visit the POQ website.
9
Table 3 also shows some interesting heterogeneity across issues and parties. Republicans exhibit the most asymmetry for the public financing issue, while Democrats exhibit the most asymmetry for the taxes and immigration issues. The anomalous results for the public financing issue could be due to John McCain’s role in campaign finance reform legislation. While such patterns are intriguing, exploring them is beyond the scope of this project, though they are an important topic for future study.
10
For example, perhaps the percentage of partisans who hold extreme views on the issue influences perceptions of where other partisans stand, or perhaps when more partisans are out of step with their party’s position, the positions of others are seen as more moderate. Our data showed little support for these hypotheses. However, because only four issues were measured, we cannot make strong or reliable inferences. Future work should explore these types of hypotheses more fully.