When election expectations fail: Polarized perceptions of election legitimacy increase with accumulating evidence of election outcomes and with polarized media.

The present study, conducted immediately after the 2020 presidential election in the United States, examined whether Democrats' and Republicans' polarized assessments of election legitimacy increased over time. In a naturalistic survey experiment, people (N = 1,236) were randomly surveyed either during the week following Election Day, with votes cast but the outcome unknown, or during the following week, after President Joseph Biden was widely declared the winner. The design unconfounded the election outcome announcement from the vote itself, allowing more precise testing of predictions derived from cognitive dissonance theory. As predicted, perceived election legitimacy increased among Democrats, from the first to the second week following Election Day, as their expected Biden win was confirmed, whereas perceived election legitimacy decreased among Republicans as their expected President Trump win was disconfirmed. From the first to the second week following Election Day, Republicans reported stronger negative emotions and weaker positive emotions while Democrats reported stronger positive emotions and weaker negative emotions. The polarized perceptions of election legitimacy were correlated with the tendencies to trust and consume polarized media. Consumption of Fox News was associated with lowered perceptions of election legitimacy over time whereas consumption of other outlets was associated with higher perceptions of election legitimacy over time. Discussion centers on the role of the media in the experience of cognitive dissonance and the implications of polarized perceptions of election legitimacy for psychology, political science, and the future of democratic society.

Introduction

Healthy democracies rest on shared confidence in election legitimacy [1]. Citizens need to regard elections as fair and legitimate for governmental effectiveness. Particularly important, yet also particularly challenging, is that people whose preferred candidate lost an election nevertheless accept the outcome as legitimate [2-4]. “Loser’s consent” is an indicator of a well-functioning democracy. Without such consent, widespread questioning of election legitimacy may reduce trust in government, catalyze mass protest, and trigger violence [5]. Americans witnessed all three outcomes following the 2020 election of President Joseph Biden and the defeat of former President Donald Trump. What does social psychology suggest about the public’s perception of election legitimacy?

We take our theoretical inspiration for this study, a naturalistic experiment during the 2020 presidential election in the United States, from classic research in social psychology and the research tradition spurred by cognitive dissonance theory [6, 7]. A recent evaluation of Leon Festinger and cognitive dissonance theory’s contributions to social psychology and society written by Lee Ross [8] makes several historical and theoretical points that anticipated our study and highlights its relevance. Ross notes a dearth of research inspired by cognitive dissonance theory that examines people’s reactions to real-world events where strong feelings of dissonance are widely experienced. A limitation of the traditional dissonance methodology, Ross notes, was that “the levels of dissonance experienced by the participants in most studies (generally college undergraduates or children) were much lower than the levels experienced by individuals who had faced soul-challenging decisions… (p. 9)” A tightly contested election, one where each side had strongly held, identity-intertwined cognitions and viewed it as a “soul-challenging decision” for the nation would seem to provide a naturalistic laboratory to examine dissonance phenomena. Ross writes:

… real-world field research received little attention. One such research target would have been assessments of the merits of political candidates immediately after versus before voting, or before versus after election results become known… However such phenomena seem to have been left to political scientists. [8, p. 10, emphasis added]

Indeed, political scientists have examined the phenomenon whereby partisans from the losing party perceive elections as less legitimate than partisans from the winning party [9-12]. Although political scientists have linked polarized perceptions of election legitimacy to theories such as cognitive dissonance theory [6, 13–17], the methods tend to confound before versus after voting with before versus after results become known because those two factors are closely intertwined in most elections.

The experience of a national election consists of a torrent of information and outcomes for people considering a consequential choice: the candidate for whom they voted. Before an election, both sides hope for and often expect their candidate to win. As election outcomes become known, dissonance can be evoked by conflicting cognitions people feel in the wake of a national election, such as the discrepancy between one’s view that that “my country is a decent place” and “my country elected a horrible person.”

In this paper, we focus on the discrepant feelings people experience during and after the election outcome as it relates to their preferred candidate. For those on the losing side, the negative outcome and unexpected disappointment stemming from their candidate’s loss is dissonant with their perceptions of the candidate’s positive attributes, which were insufficient to compel a winning majority. One way this can be accomplished is by questioning the legitimacy of votes, suggesting that their candidate’s loss is illegitimate and that, if votes had been counted correctly, the candidate would not have lost. For winners, the expected positive outcome of their candidate’s win is dissonant with any lingering doubts about their candidate’s ability to govern effectively and other perceptions of the candidate’s weakness. Reducing this dissonance can be done by affirming the legitimacy of votes, thereby bolstering the candidate’s widespread appeal and legitimacy of the candidate’s win. Theories of dissonance-induced rationalization thus imply that polarized perceptions of election legitimacy should increase over time as the election outcomes become widely known [18].

We report the results of a naturalistic survey experiment conducted during the 2020 presidential election in the United States. We tested whether Democrats, whose presidential candidate Joseph Biden won the election, would perceive the election outcome as more legitimate, as indicated by confidence that votes were correctly counted, than Republicans, whose candidate Donald Trump lost the election. We also tested the crucial prediction that these polarized differences would increase over the two weeks following Election Day, as votes were counted, outcome certainty increased, and President Biden was widely declared the winner.

The present study takes advantage of the unique circumstance of the 2020 presidential election in the United States to isolate knowledge of election outcome from the possibility of voting or influencing votes. The present study also provides indirect correlational evidence for the role of social confirmation through the consumption of polarized media in reducing dissonance by increasing polarized perceptions of election legitimacy. We do so by examining whether trust and consumption of media outlets that voiced skepticism about election legitimacy (i.e., Fox News) was correlated with decreased perceptions of election legitimacy as vote counts and the election outcome became more widely known.

Polarized perceptions of election legitimacy

Previous research in political science demonstrates that the tendency for winners to perceive elections as more legitimate than losers increases from before to after elections, consistent with predictions derived from cognitive dissonance theory [18]. However, those studies necessarily confound whether people have learned the election outcome with the possibility of voting in the election [9-12], leading to theoretical imprecision in what drives polarized perceptions of election legitimacy. Before an election, citizens can still influence election outcomes by voting and convincing others to vote for their candidate. After an election, citizens have learned the outcome, typically announced the evening of Election Day in the United States, and they can no longer cast or influence votes. Of course, as the unfolding of the 2020 American presidential election revealed, people could also question the legitimacy of the voting process after its conclusion through a multitude of actions such as legal disputes about election practices or disrupting the certification of voting outcomes. Nevertheless, there is a qualitative difference between these highly unusual actions and the sanctioned actions people can take to influence voting before an election day. Pre- to post-election comparisons that have been typically featured in research confound two plausible causes of polarized perceptions: outcome knowledge and opportunities to influence elections.

Disentangling this confound is important because the mere act of voting may increase polarized attitudes [19-21]. In one study, voters entering a polling station, who had not yet voted, held less favorable opinions of their candidate than voters who were exiting the polling station, having just cast their vote [19]. Without knowing the election outcome, the act of voting was sufficient to increase polarized perceptions of the candidate. This raises the possibility that increasingly polarized perceptions of election legitimacy from pre- to post-election is due to people having voted, not to their having learned the election outcome.

The 2020 U.S. presidential election presented an opportunity to disentangle these interpretations. The election had a historically high turnout, an unprecedented number of mail-in and absentee ballots, and social distancing practices during the COVID-19 pandemic. This contributed to a period following Election Day when votes were cast but the winner was uncertain. Four days after Election Day, on November 7, 2020, Biden was widely declared the winner (Fig 1). Comparing perceptions of election legitimacy during the contested, undeclared period with perceptions after the outcome was widely declared removes the possible impact of individuals influencing the actual vote total (by either voting or encouraging others to vote), thus affording greater theoretical precision in testing the influence of learning the election outcome. With widespread predictions that the 2020 U.S. presidential election would not be called on election night, we designed a study to disambiguate these interpretations, and moreover, to examine correlated emotions and the correlated consumption of and trust in polarized media.

Fig 1

Timeline of election events over course of data collection.

Emotional election outcomes

We also examined whether the psychological dynamics of a presidential election would provoke associated positive and negative emotions. Because partisans are personally invested in election outcomes, their group-based emotional experience should reflect emerging knowledge that their favored candidate won or lost the election [22, 23]. Negative feelings of anger, anxiety, irritability, and nervousness are indicators of disappointment and, we reason, the discomfort stemming from any cognitive dissonance elicited from an unexpected election loss [24]. Positive emotions of hope, happiness, and excitement are indicators of the pleasure of a hoped-for but uncertain election win. Given that President Biden won the election, we examined whether Democrats would exhibit subsiding negative emotions while Republicans would exhibit intensifying negative emotions, with positive emotions following the reverse pattern.

Polarized media, social confirmation, and election legitimacy

Elections are social events, so it is important to also examine dissonance reduction processes within the broader social context that occurs during and after elections. The arousal and reduction of cognitive dissonance is often a social process [7]. Yet much research inspired by cognitive dissonance theory has focused on the internal dynamics and attitudinal consequences of having engaged in self-relevant, seemingly freely chosen, counter-attitudinal behavior [25]. In his essay on cognitive dissonance theory, Ross noted that “perhaps the most important shortcoming of the dissonance theory tradition was its almost exclusive focus on individual rather than collective processes” (p. 15).

The experiences of Democrats and Republicans in the wake of the 2020 presidential election as they made sense of (and rationalized) the outcome were likely influenced by the collective processes reflected by partisan media. Major media outlets and their associated information ecosystems both convey information about the reactions of fellow Democrats and Republicans, and provide rationalizing information consistent with dissonance reducing claims about the election [26, 27]. Partisan media may serve the social function of reflecting and abetting polarized beliefs [28-31].

Reducing cognitive dissonance is one means towards restoring and maintaining cognitive consistency [32-34]. The influence of partisan media in polarizing perceptions of election legitimacy should therefore depend on engagement with consistent, homogeneous media that provide coherent, confirmatory information. Consistent with this analysis, liberals and conservatives with more extreme stances exhibit greater homogeneity in their social media networks [35], which may similarly occur with mainstream media outlets.

In the context of U.S. politics, Fox News is singularly trusted by Republicans and distrusted by Democrats and Independents [36, 37]. Unlike other major mainstream media outlets, Fox News stood out in questioning the legitimacy of the 2020 U.S. presidential election, asking whether votes were correctly counted or fraudulent [38, 39]. In this way, partisan media may provide evidence to support individuals’ polarized perceptions of election legitimacy, even in the face of accumulating evidence confirming vote counts and the election outcome. We thus examined whether increased polarized perceptions of election legitimacy followed engagement with partisan media. We specifically examined whether trust in and consumption of mainstream media outlets was associated with increased perceptions of election legitimacy whereas trust and consumption of Fox News was associated with reduced perceptions of election legitimacy.

Study overview

We examined polarized perceptions of election legitimacy in a naturalistic experiment following the 2020 U.S. presidential election. Because it was widely anticipated that the election would not be called on election night, we randomly assigned participants from a larger sample that had participated in earlier research related to perceptions of COVID-19 policies and the media, to one of two conditions: People either completed a survey study during the week following Election Day, with votes cast but the outcome still to be determined, or the following week, when Biden was the widely declared winner (Fig 1). We defined the week following Election Day (November 4–8) as “Undeclared” because many states’ vote counts had not been confirmed and the outcome was unknown by everyone, particularly during the days immediately following election day. We defined the following week (November 9–15) as “Declared” because all states’ vote counts had been confirmed, the Associated Press (and most other media outlets, including Fox News, if not its prominent commentators) had declared President Joe Biden as winner, and he had delivered his acceptance speech. The time frame covered by our study obviously reflected a continuous process of new information and events. Nevertheless, the dividing line reflects a period when no one knew who was declared winner versus a period when most people presumably knew that Biden was declared winner.

Thus, the design was not a cross-sectional design whereby participants were selected at two time points, but a true experiment, where participants had an equal chance of being either in the “Undeclared” or the “Declared” period. The study, however, has elements of a natural experiment, in that the independent variable (“Undeclared” vs. “Declared” election outcome) depended on events occurring in the world outside of the researchers’ control: When the victor of the 2020 U.S. presidential election was publicly declared by the major media outlets.

We used an established measured of perceived election legitimacy, operationalized as confidence that votes were counted as intended [3, 10]. We also measured participants’ expectations about who would win the election. We examined whether Democrats and Republicans would expect their candidate to win, whether Democrats would perceive the election as more legitimate than Republicans, and whether these polarized perceptions of election legitimacy would increase over time.

We also measured people’s emotional reactions to the election. From the Undeclared to Declared periods, we examined whether Democrats’ negative emotions would subside while positive emotions would increase, and whether Republicans’ negative emotions would increase while positive emotions would subside. We further examined the associations between these emotional reactions and perceptions of election legitimacy.

Finally, we asked participants to report how much they trusted and consumed Fox News and 14 other widely consumed media outlets including CNN, The New York Times, and The Wall Street Journal. Confirming other research, we expected Republicans to trust and consume Fox News more than Democrats, who would trust and consume other media outlets [37, 40]. We explored whether trust and consumption of Fox News and of other media would independently predict and moderate perceptions of election legitimacy and changes over the two-week period.

Method

The Institutional Review Board at the University of Colorado categorized the study as Exempt (Protocol 20–0197). All data, materials, and code are available at https://osf.io/ewr7g/.

Participants

Participants were U.S. residents (N = 1,236, 44.5% female, 47.2% male, with the remaining other/unspecified), recruited using ROI Rocket, and paid $4. The sample was diverse in age (Mage = 49.71; SDage = 15.21, range [18, 89]) and ethnicity (11.7% African American, 8.2% Asian American, 6.8% Latin American, 62.7% White, and 11.8% other or declined to provide). As in previous research [e.g., 41], we measured partisan identification using a two-step procedure with dichotomous questions from the American National Election Study. This allowed categorization, including leaners, of participants as Democrat (N = 565), Republican (N = 466), or Independent (N = 194), with 11 who did not report their partisan identification.

The sample yielded an 80% chance to detect an effect of F = 3.0 and the sample size was determined based on available funding. Sample sizes differ across analyses due to missing data.

Procedure

A total of 1,672 potential participants were randomly assigned to one of two survey periods, either November 4–8 or November 9–15, 2020. Response rates for both periods (583 of 836 = 70% during the Undeclared period; 653 of 836 = 78% during the Declared period) were consistent with online panel response rates [42] and were not significantly different from each other (p = .134). We referred to the first period as the “Undeclared” period because all votes had been cast but not fully counted leaving the election outcome unknown and not declared.

Anticipating a delay between election night and declared election results, we planned to leave the Undeclared group’s survey open until a winner was officially called; we closed the first group’s survey on November 8th, one day after the Associated Press called the election for Biden. We then launched the survey for the second, Declared group on November 9th. We referred to the second period as “Declared” because Joe Biden had delivered his acceptance speech and had been widely declared to have won the election. Of our Undeclared sample (N = 583), 17 people participated during the Declared period (because they participated on November 8th). Excluding these people does not substantively impact the pattern of results (see OSF); of course, including them as we did works against the hypotheses.

Measures

Expected outcome

We measured participants’ expected election outcome with a single item, with wording in brackets for participants in the declared period: “Who do [did] you think will [would] ultimately win the presidential election?” Responses were on an ordinal scale presented without numbers, with counterbalanced scale anchors (1 = Definitely Donald Trump; 5 = Unsure/Toss-up; 9 = Definitely Joe Biden).

Perceived election legitimacy

We measured participants’ perceived election legitimacy by averaging answers to two questions about confidence that votes were counted correctly. One item measured perceptions of own vote confidence and the other perceptions of nationwide vote confidence, respectively: “How confident are you that your vote in the general election was counted as you intended?” and “How confident are you that votes nationwide in the general election were counted as voters intended?” [3, 10]. Participants answered on two ordinal scales presented without numbers (1 = Not at all confident; 5 = Very confident). Responses were highly correlated (r = .75).

Emotion

Participants reported their emotions about the presidential election: “When you think about the election, how much do you feel each of the following?” (1 = Not at all; 7 = Extremely). Negative emotions were averaged into an index comprising anger, guilt, shame, embarrassment, nervousness, distress, and irritability (α = .88). Positive emotions were averaged into an index comprising pride, gratitude, hope, happiness, and excitement (α = .93).

Media trust and consumption

We measured participants’ trust and consumption of media by asking people how much they trusted and consumed 15 media outlets: Fox News, ABC News, AOL News, CBS News, CNN, Huffington Post, MSNBC, NBC News, NPR, New York Times, PBS, USA Today, Washington Post, Wall Street Journal, and Yahoo News. We did not include a broader range of conservative media outlets such as One American News Network, Newsmax, or Breitbart because they lacked available and established full-text databases that we required for a project unrelated to the present manuscript regarding linguistic analysis of media content. Participants answered, “How much do you distrust or trust the accuracy of reporting for that source?” (1 = Distrust completely; 5 = Trust completely). We also asked them, as part of the larger research project on media consumption and COVID-19, “In general, how much do you get news about COVID-19 from each source?” (1 = Not at all; 5 = A great deal). Participants were presented with non-numeric scales when answering these items. For each participant, we computed the average correlation between trust and consumption ratings for each of the 15 media outlets. Within participants, the average r was .51 (SD = 0.33). Examined differently, for each media outlet, we calculated the correlation across participants of their trust and consumption ratings (the average r across 15 outlets was also equal to .51, range [.36, .67]). Patterns of significance are the same when we separately examine the measures of trust and consumption. Because previous research demonstrated that Fox News is uniquely trusted by Republicans compared with other media outlets [37, 40], we averaged participants’ ratings of the 14 media outlets other than Fox News.

Results

Expected election outcome

Democrats and Republicans had different expectations of the election outcome, as reflected by a main effect of partisanship in a 3(partisan identification: Democrat, Republican, or Independent) × 2(timing: Undeclared, Declared) ANOVA (F(2, 1078) = 284.76, p < .001, = .359). Compared with the scale midpoint of 5, Republicans expected Trump to win (M = 3.63, SD = 2.64; F(1, 1078) = 143.35; p < .001, = .117), Democrats expected Biden to win (M = 7.42, SD = 2.02; F(1, 1078) = 484.57, p < .001, = .310), with Independents in between (M = 5.38, SD = 2.30; F(1, 1078) = 2.82, p = .093, = .004). The interaction between partisan identification and timing was not significant (F(2, 1078) = 2.20, p = .112, = .004). Each side believed their candidate would win, meaning that after Biden was declared winner, Democrats’ expectations were confirmed, and Republicans’ expectations were disconfirmed.

Perceptions of election legitimacy

Our central prediction was that polarized perceptions of election legitimacy would increase after Biden was the widely declared winner compared with the undeclared period when votes were cast but the outcome was unknown (Fig 2). To test this, we conducted a 3(partisan identification: Democrat, Republican, Independent) × 2(timing: Undeclared, Declared) ANOVA on the aggregate measure of perceived election legitimacy. The key predicted interaction between partisan identification and timing was significant (F(2, 1202) = 10.80, p < .001, = .018). From the Undeclared to Declared periods, Democrats became more confident that votes were counted as intended (MUndeclared = 3.92, SDUndeclared = 1.00; MDeclared = 4.37, SDDeclared = 0.92; F(1, 1202) = 21.18, p < .001, = .017), while Republicans became less confident that votes were counted as intended (MUndeclared = 2.99, SDUndeclared = 1.18; MDeclared = 2.79, SDDeclared = 1.24; F(1, 1202) = 3.94, p = .047, = .003), and Independents’ confidence remained unchanged over time (MUndeclared = 2.96, SDUndeclared = 1.28; MDeclared = 3.02, SDDeclared = 1.45; F(1, 1202) < 0.01, p = .983, < .001).

Fig 2

Violin density plots, means, and +/–SE of perceived election legitimacy among Democrats, Republicans, and Independents who reported confidence that both their own and nationwide votes were counted as intended.

The width of the density plot represents the relative portion of each sample at each value of election legitimacy perception.

The increase in Democrats’ perception of election legitimacy was larger in magnitude than the decrease in Republicans’ perception of election legitimacy. This may simply reflect that Democrats’ increasing confidence that votes were counted correctly coincided with accumulating evidence from state certification of vote counts. Republicans’ decreasing confidence that votes were counted as intended occurred despite official certification of vote counts.

Emotions

The polarized perceptions of election legitimacy and expectations that were dashed or confirmed were accompanied by polarized emotions (Fig 3). A 3(partisan identification: Democrat, Republican, Independent) × 2(timing: Undeclared, Declared) × 2(emotion valence: negative, positive) ANOVA with repeated measures on the last factor revealed a 3-way interaction (F(1, 1208) = 87.05, p < .001, = .067).

Fig 3

Violin density plots, means, and +/–SE of average emotion, negative in the top panel and positive in the bottom panel, among Democrat, Republican, and Independent participants during two weeks during and after the 2020 presidential election.

Participants reported their negative emotions (anger, shame, embarrassment, nervousness, distress, and irritability) and positive emotions (pride, gratitude, hope, happiness, and excitement). The width of the density plot represents the relative portion of each sample at each value of negative and positive emotion.

For negative emotions, there was a significant interaction in a 3(partisan identification: Democrat, Republican, Independent) × 2(timing: Undeclared, Declared) ANOVA (F(2, 1208) = 12.83, p < .001, = .013). From the Undeclared to Declared periods, Republicans’ negative emotions increased (MUndeclared = 2.93, SDUndeclared = 1.39, MDeclared = 3.19, SDDeclared = 1.54; F(1, 1208) = 3.92, p = .048, = .003) while Democrats’ negative emotions decreased (MUndeclared = 3.35, SDUndeclared = 1.62, MDeclared = 2.71, SDDeclared = 1.46; F(1, 1208) = 27.88, p < .001, = .023) and Independents’ negative emotions were unchanged (MUndeclared = 2.64, SDUndeclared = 1.54, MDeclared = 2.54, SDDeclared = 1.47; F(1, 1208) = 0.18, p = .667, < .001). Positive emotions revealed the inverse shift, as reflected by an analogous interaction (F(2, 1208) = 40.43, p < .001, = .132). From the Undeclared to Declared periods, Democrats’ positive emotions increased (MUndeclared = 3.26, SDUndeclared = 1.60, MDeclared = 4.66, SDDeclared = 1.75; F(1, 1208) = 83.54, p < .001, = .065) while Republicans’ positive emotions decreased (MUndeclared = 3.11, SDUndeclared = 1.65, MDeclared = 2.61, SDDeclared = 1.73; F(1, 1208) = 11.83, p = .0006, = .010) and Independents remain unchanged (MUndeclared = 2.44, SDUndeclared = 1.47, MDeclared = 2.25, SDDeclared = 1.60; F(1, 1208) = 0.24, p = .628, < .001).

These emotional reactions were associated with perceptions of election legitimacy. In a multiple regression analysis, negative emotions were associated with lower perceptions of election legitimacy (b = –0.12, F(1, 1200) = 30.03, p < .001, = .024) whereas positive emotions were associated higher perceived election legitimacy (b = 0.19, F(1, 1200) = 105.70, p < .001, = .081), controlling for partisan identification (F(2, 1200) = 122.20, p < .001, = .169), timing (b = 0.02, F(1, 1200) = 0.07, p = .799, < .001), and their interaction (F(2, 1200) = 0.94, p = .390, = .002). To the extent that people perceived the election as legitimate, they reported more positive emotion and less negative emotion.

Media trust and consumption

Democrats and Republicans were polarized in their trust and consumption of media outlets (Fig 4). A 3(partisan identification: Democrat, Republican, Independent) × 2 (election timing: Undeclared vs. Declared) × 2(media outlet: Fox News, Other Outlet) ANOVA with repeated measures on media outlet revealed a nonsignificant interaction between partisan identification and media outlet (F(2, 1206) = 37.00, p < .001, = .329). However, Republicans reported trusting and consuming Fox News (M = 2.96, SD = 1.28) more than did Democrats (M = 1.97, SD = 1.10, F(1, 1209) = 185.91, p < .001, = .133), who reported trusting and consuming the 14 other outlets (M = 2.89, SD = 0.66) more than did Republicans (M = 1.99, SD = 0.84; F(1, 1209) = 357.59, p < .001, = .228). Independents’ ratings of media outlets were in between Democrats and Republicans. The 3-way interaction was not significant (F(2, 1206) = 1.03, p = .356, = .002), indicating that this pattern of results did not change significantly between the undeclared and declared periods. These results confirm that Democrats and Republicans trust and consume different media outlets, with Fox News being uniquely highly rated by Republicans.

Fig 4

Mean and–/+ SE of media trust and consumption of 15 major media outlets, the average of participants’ trust in each outlet and their reported consumption of COVID-19 news from each outlet.

To explore these relationships more extensively, we regressed the measure of perceived election legitimacy on participant partisan identification, election timing, ratings of Fox News, ratings of other media outlets, and their interactions (Table 1). Consistent with hypotheses, to the extent that participants trusted and consumed Fox News, they perceived lower election legitimacy (main effect of Fox News ratings: b = –0.20), that decreased over time (election timing × Fox News ratings: b = –0.17). To the extent participants trusted and consumed other outlets, they perceived higher election legitimacy (main effect of other outlet consumption: b = 0.40) that increased over time (election timing × other outlets ratings: b = 0.28). Consumption and trust of polarized media predicted changes over time in perceptions of election legitimacy.

Table 1

Multiple regression predicting perceived election legitimacy from participant partisan identification, election timing, and their interaction (Model 1) and additionally from the mean centered averages of perception of Fox News and of other media outlets, and their interactions (Model 2).

	Model 1					Model 2
Predictor	b	SE	t	p	b		SE	t	p
Partisan Identification
Democrat vs. Republican	–1.26	0.07	–17.66	< .001	–0.97		0.09	–10.46	< .001
Independent vs. Democrat and Republican	0.55	0.09	6.11	< .001	0.54		0.10	5.28	< .001
Election Timing
Election timing	0.08	0.07	1.16	.248	0.15		0.09	1.69	.091
Media Measures
Fox News trust and consumption	–	–	–	–	–0.20		0.04	–5.07	< .001
Other media trust and consumption	–	–	–	–	0.40		0.05	7.66	< .001
Partisan Identification x Election Timing Interactions
Democrat vs. Republican x Election timing	–0.65	0.14	–4.58	< .001	–0.27		0.19	–1.48	.140
Independent vs. Democrat and Republican x Election timing	0.10	0.18	0.56	.574	0.11		0.21	0.51	.607
Media Measures x Election Timing Interactions
Fox News trust and consumption x Election timing	–	–	–	–	–0.17		0.08	–2.20	.028
Other media trust and consumption x Election timing	–	–	–	–	0.28		0.10	2.71	.007
Partisan Identification x Media Measures x Election Timing Interactions
Democrat vs. Republican x Election timing x Fox trust and consumption	–	–	–	–	0.22		0.14	1.61	.109
Independent vs. Democrat and Republican x Election timing x Fox trust and consumption	–	–	–	–	–0.02		0.20	–0.11	.915
Democrat vs. Republican x Election timing x Other media trust and consumption	–	–	–	–	0.59		0.22	2.62	.009
Independent vs. Democrat and Republican x Election timing x Other media trust and consumption	–	–	–	–	–0.09		0.24	–0.38	.704
Partisan Identification x Both Media Measures x Election Timing Interaction
Democrat vs. Republican x Election timing x Fox trust and consumption x Other media trust and consumption	–	–	–	–	–0.29		0.14	–2.10	.036
Independent vs. Democrat and Republican x Election timing x Fox trust and consumption x Other media trust and consumption	–	–	–	–	–0.20		0.17	–1.21	.226

Note: Model 1: R2 = 0.248, df = 1202; Model 2: R2 = 0.312, df = 1184. Regressions included two contrast coded predictors: One compared Democrats (–1/2) with Republicans (+1/2), and one compared Independents (–2/3) with Democrats and Republicans (+1/3 for both).

Moreover, the analysis yielded a 4-way interaction (partisan identification × election timing × other outlets ratings × Fox News ratings: b = –0.29). We decomposed the interaction by partisan identification. Among Democrats, trust and consumption of Fox News, independent of ratings of other outlets, was associated with reductions in the increase of perceptions of election legitimacy over time (election timing × Fox News ratings: b = –0.29, t = –3.04, p = 0.002). Democratic ratings of other outlets were not associated with changes over time in perceived election legitimacy (election timing × Other Outlet ratings: b = –0.04, t = –0.42, p = .676). Among Republicans, in contrast, trust and consumption of other media outlets, independent of Fox News ratings, was associated with a reduced decline in perceived election legitimacy (election timing × Other Outlet ratings: b = 0.54, t = 2.89, p = .040). Higher Republican ratings of Fox News were not associated with changes over time in perceived election legitimacy (election timing × Fox News ratings: b = –0.07, t = –0.74, p = .458).

Thus, to the extent that participants trusted and consumed news outlets that are typically distrusted and avoided by their political in-group—Fox News for Democrats and other sources for Republicans—their perceptions of election legitimacy followed the prevailing pattern within their political group less strongly. We did not, however, find that trusting and consuming news outlets that are typically aligned with political in-group—other outlets for Democrats and Fox News for Republicans—were associated with stronger prevailing patterns within their in-group. We are hesitant to draw overly strong conclusions based on an exploratory analysis of higher order interactions. Yet the results are consistent with the possibility that trust and consumption of less ideologically diverse media outlets is associated with greater polarized perceptions of election legitimacy.

Perceived election legitimacy of own vote and nationwide vote

Past findings suggest that people trust their local government more than they trust the national government [45]. This raises the possibility that people who lose elections lose faith in the national election legitimacy because they suspect that other people’s (i.e., those from other areas) votes are not counted as intended. In the perceived election legitimacy analysis reported above, there was a significant 3-way interaction between partisan identification, timing, and vote type (F(2, 1201) = 8.71, p < .001, = .014). The partisanship × timing interaction was larger for perceived legitimacy of nationwide (F(2, 1201) = 17.89, p < .001, = .029) than for own votes (F(2, 1202) = 3.60, p = .027, = .006). From the Undeclared to Declared period, Democrats’ perceived legitimacy increased for both nationwide votes (MUndeclared = 3.78, SDUndeclared = 1.10, MDeclared = 4.36, SDDeclared = 0.94; F(1, 1201) = 33.18, p < .001, = .027) and own votes (MUndeclared = 4.07, SDUndeclared = 1.03, MDeclared = 4.37, SDDeclared = 0.97; F(1, 1202) = 8.01, p = .005, = .007). Among Republicans, from the Undeclared to the Declared period, the decrease in perceived legitimacy of nationwide votes was larger (MUndeclared = 2.64, SDUndeclared = 1.25, MDeclared = 2.33, SDDeclared = 1.33; F(1, 1201) = 7.56, p = .006, = .006) than for own votes, which did not significantly decrease (MUndeclared = 3.35, SDUndeclared = 1.34, MDeclared = 3.24, SDDeclared = 1.44; F(1, 1202) = 0.93, p = .335, = .001). The differential decline for Republicans may reflect that people know more about their own votes than they do about others’ votes. For their own votes, people have direct, personal experience voting and trust in their local district’s counting process whereas the uncertainty and lack of experience with nationwide votes may lend itself more to rationalizing processes. An analogous difference among Democrats’ between confidence in one’s own vote versus nationwide votes may not have emerged because of a ceiling effect for confidence.

Discussion

Citizens in healthy democracies view votes as legitimate even when their preferred candidate loses [2-4]. Judging elections as fair and legitimate is critical for governments to operate effectively [4, 43]. In contrast with these democratic ideals, election losers tend to perceive elections as less legitimate than do winners [10, 18]. Researchers have attributed such outcome-dependent perceptions of election legitimacy to the reduction of cognitive dissonance wherein losers perceive elections as less legitimate, with the implication that in a legitimate election their preferred candidate would have won more votes [18]. Consistent with this reasoning, polarized perceptions of election legitimacy increase from pre- to post-election. In previous research, however, knowledge of the outcome from pre- to post-election was confounded with the possibility of voting or influencing votes, which is itself sufficient to polarize election attitudes. The present study used a naturalistic experiment to demonstrate that learning the election outcome increases polarized perceptions of election legitimacy, even when it is no longer possible to exert electoral influence via voting, consistent with theories of cognitive dissonance.

In the context of the U.S. 2020 presidential election, Democrats (winners) were more confident than were Republicans (losers) that both their own and nationwide votes were counted correctly. These polarized perceptions increased from Election Day through the second week following Election Day as evidence accumulated that President Joe Biden won. This design removes the possibility of influencing votes, more precisely implicating knowledge of the election outcome in provoking dissonance. As more votes were counted and Democrats’ expected Biden win was confirmed, they became more confident that the votes were counted correctly. Over the same period, as more votes were counted and Republicans’ expected Trump win was disconfirmed, they became less confident that votes were counted correctly. Along with perceptions of election legitimacy, partisans’ emotions became increasingly polarized. From the undeclared to the declared period, Democrats’ emotions became more prevailingly positive and less negative while Republicans’ emotions became less positive and more prevailingly negative. These emotional profiles may both reflect the arousal of dissonance [22-24] and increasingly confident assessments that the election was illegitimate or legitimate.

These polarized assessments of electoral integrity and emotions corresponded with polarized media. Trust and consumption of Fox News independently predicted lower perceived election legitimacy that decreased over time. Trust and consumption of other media outlets independently predicted higher perceived election legitimacy that increased over time. These findings are consistent with the possibility that polarized media functions as social confirmation that polarizes perceptions of election legitimacy [26, 28, 38, 44].

The present findings illustrate a point made by Eliot Aronson and Carol Tavris [45], that social confirmation by groups with whom one is strongly connected facilitates dissonance-reducing rationalizations that can distort evidence, or as they put it: “[W]hen people feel a strong connection to a political party, leader, ideology, or belief, they are more likely to let that allegiance do their thinking for them and distort or ignore the evidence that challenges those loyalties” [45].

It is noteworthy that essays by both Lee Ross [8] and Eliot Aronson [45], two prominent social psychologists reflecting on the relevance of cognitive dissonance theory to contemporary issues, focus on role of the collective in fomenting rationalization. Although this was not the focus of much cognitive dissonance research, which emphasized intraindividual rather than collective processes, their observations resonate with one of the classic studies conducted by Festinger and colleagues. In their study of the reactions of a doomsday cult whose prophesied alien arrival failed to materialize, Festinger and colleagues explained how social processes enable belief persistence when confronted with disconfirming evidence [7, p. 4]:

The individual believer must have social support. It is unlikely that one isolated believer could withstand the kind of disconfirming evidence we have specified. If, however, the believer is a member of a group of convinced persons who can support one another, we would expect the belief to be maintained…

In other words, reducing cognitive dissonance to maintain disconfirmed expectations is not simply an individual rationalization process, but is bolstered by social support, justifications, and the provision of rationalizing information by others. Although media outlets are different from tightly knit social groups like cults, media may serve a similar function by providing polarized information that helps protect and confirm beliefs shaken by disconfirming evidence, processes hinted at by our findings regarding trust and consumption of Fox News and other mainstream media.

Fox News was uniquely trusted and consumed by Republicans and not Democrats, but this deserves further comment. As noted earlier, our selection of media outlets reflected the availability of searchable full text databases of media content for use in a different project. We also suspected that the Wall Street Journal would be trusted and consumed by Republicans more than Democrats, and that outlets like USA Today and traditional network television would be rated equally by Democrats and Republicans. But this was not the case. Right-wing media outlets have increased in recent years, as Republicans became less trusting of so-called “mainstream media,” which President Trump had referred to as “fake news” [43, 44]. An important task for future research will be to study the number of outlets trusted and consumed by Republicans.

Future research might also pursue several additional clarifying questions. One would be to differentiate the influence of polarized media from that of political leaders such as President Joe Biden and President Donald Trump. Another would be to investigate why people have more confidence that their own votes were counted correctly than that nationwide votes were counted correctly, much as they trust their local government more than they trust the national government [46]. People may have more direct knowledge that their ballot was clearly completed and mailed or deposited in the ballot box, whereas their knowledge of nationwide ballots are indirect, distant, and open to suspicion or reinterpretation. Examining an expanded time course of polarized perceptions of election legitimacy is another important question for future work. The present study was limited to the two weeks following Election Day so cannot directly address questions about the persistence of polarized perceptions, as would be implied by cognitive dissonance theory. However, evidence from national polls indicate that polarization persisted until late January 2021, nearly three months after the election [47]. And, of course, the January 6, 2021, attack on the U.S. Capitol suggests that persistent assessment of an illegitimate election was sufficient to motivate acts of political violence [48].

Notwithstanding these open questions, the present findings have implications for understanding and improving media engagement to reduce polarized perceptions of election legitimacy. Others have noted how Fox News undermines perceptions of election legitimacy [37, 40]. The present results additionally undergird the importance of consuming ideologically diverse media sources to mitigate polarized perceptions of election legitimacy, a theme worth emphasizing in civic education.

Pursuing these topics is important because they inform understanding of both psychology and democracy. A recent analysis found that perceived electoral integrity is less conditional on electoral outcomes in healthy democracies. Healthier democracies not only evince less of a negative effect on perceived electoral integrity among the losing party but also less of an increase among the wining party [49]. In the U.S. 2020 presidential election, as evidence of an election’s outcome accumulated, vote confidence was conditional on knowledge of the election outcome in both directions—winners became more confident that votes were correctly counted whereas losers became less confident. Social confirmation abets the dissonance reducing process for winners and losers alike, illustrating how politically polarized vote confidence is both fomented and reflected by polarized media information ecosystems. And cognitive dissonance theory, when explored in the context of contemporary events of dramatic consequence, continues to yield insights and inspire new research questions.

10 Aug 2021

PONE-D-21-23399

When Election Expectations Fail:

Polarized Perceptions of Election Legitimacy Increase with Accumulating Evidence of Election Outcomes and with Polarized Media

PLOS ONE

Dear Dr. Van Boven,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

I asked two experts in the field to review this manuscript. I synthesize their key comments with my own (based on an independent reading) below. Overall, all three of us think that this paper should be published in PLOS ONE, though we differ somewhat in how much more work is needed.

The two places that I think need to be revised are the positioning and (at least one of) the analyses. For the positioning, you lean heavily on dissonance theory, which certainly makes sense intuitively, and yet as R2 points out, there are other plausible explanations for why belief in the election’s legitimacy could change with time. Beyond that, there is no direct measure of dissonance, making it that much more difficult to know if dissonance is driving this result or not. I would suggest significantly reducing the exposition on dissonance and alluding to the entire idea as just a possibility, rather than a theory to build on (with these results being confirmation of that theory). For the analyses, I agree with R2’s belief that the polarization effect should (based on your set up) be moderated by in-group media consumption. I encourage you to run the set of analyses that the reviewer suggested. If the results don't support your theorizing, I would suggest that you both downplay the media consumption angle entirely AND report the lack of a result in an appendix.

Aside from these two larger issues, I suggest you carefully read and respond to the other comments made by both reviewers as they clearly took the care to provide feedback that would improve your work.

On the whole, I think that with these changes, this manuscript has a clear path to publication. Short of something significantly new in the revision, I do not anticipate sending the manuscript back out to review.

As always, please provide a detailed summary of the changes you have made in response to the entire review team’s comments.

Best of luck.

==============================

Please submit your revised manuscript by Sep 24 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Jeff Galak, PhD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

3. Please include your full ethics statement in the ‘Methods’ section of your manuscript file. In your statement, please include the full name of the IRB or ethics committee who approved or waived your study, as well as whether or not you obtained informed written or verbal consent. If consent was waived for your study, please include this information in your statement as well.

4. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I was a reviewer for this manuscript when it was previously submitted to a different journal, and I very much appreciate how the authors have altered the paper to reflect and address many of my earlier concerns. I also thought that the new material from Lee Ross nicely motivated the present investigation. What follows are some lingering issues that I think can easily be addressed in a revision:

Line 82: Unfortunately, I can’t quite parse the following sentence: “For winners, confirming knowledge that their expected and hoped-for win needs to be psychologically reconciled with any lingering doubts about the candidate’s electability and other perceptions of the candidate’s weaknesses” -- even by manipulating whether “confirming” is intended as a verb or an adjective.

Line 102: “The present study also provides evidence for the role of social confirmation in reducing dissonance through motivated reasoning about election legitimacy” – again a bit difficult to parse -- is the role played by social confirmation or motivated reasoning? I realize that “dissonance through motivated reasoning about election legitimacy” is intended to be a single unit, but maybe unpacking the sentence a bit would be helpful.

Line 119: I appreciate the new edit, but “high” should be “highly,” yes?

Line 243: I also appreciate this acknowledgement but maybe it’s also worth acknowledging somewhere that, especially in light of the handy figure, the outcome of the election was somewhat of a continuing process, not a bright-line distinction.

Line 298-303: Democrats’ conviction that their votes were counted correctly evidently increased much more than the barely significant drop in Republicans’ confidence that their votes were counted correctly, yes? Maybe worth mentioning that difference?

More importantly, why do these means differ from those reported in the next paragraph in

Lines 312-313 and Lines 317-318? No doubt I have missed some distinction, but I’d nevertheless appreciate a clarification.

Line 318-320: “This differential decline may simply reflect that people know more about their own votes than they do about others’ votes.” Perhaps, but then why don’t we see a similar increase for Democrats’ confidence in their own vote vs. others’ votes? Perhaps a ceiling effect?

Line 437: There appears to be a word missing

Line 438: Given the measures employed in the study, it’s a bit odd to refer to, for example, Fox News as a source of collective dissonance reduction in the same way that a doomsday cult (or other tightly-knit social groups) apparently was. The latter obviously offered a much more intimate connection for the target perceiver than the former, which could simply (or primarily) represent a source of biased information, as opposed to all the other ways in which fellow cult members could be a source of comfort when a prophecy is seemingly disconfirmed. Perhaps some such acknowledgment of this distinction is warranted.

Reviewer #2: I have to admit that I have never really understood PLOS ONE’s publication criteria and the role of PLOS ONE reviewers. If my task was to examine if the manuscript reports empirical research that satisfies the basic standards of science and if the description of the research is comprehensible for scientists from other fields and practitioners then my answer is “Yes this manuscript should be published in PLOS ONE.” In other words, the manuscript satisfies the 7 criteria listed in the Guidelines for Reviewers.

Below a more nuanced review that highlights the strengths and weaknesses of the manuscript. I am providing rather detailed input to give the authors the opportunity to increase the scientific contribution and impact of their paper.

There are clearly many things to like about this manuscript. The experimental manipulation is a strong point. So is the theoretical research question as well as the attempt to test theoretical predictions about cognitive dissonance in an applied setting. The study is well-conducted, and the analyses are mostly correct and reported in a straightforward manner. I liked the violin density plots. The findings provide evidence for the idea that emotions and perceptions of election legitimacy polarized over time after the 2020 presidential election.

Neither the predictions nor the results are particularly surprising. Most people, even those without any training in social sciences, would predict that losers perceive elections as less legitimate (and experience more negative emotions) than winners, and that this difference increases as the outcome of the elections become more and more clear. Once does not need cognitive dissonance to explain this effect.

The difference between the two groups, “Declared” and “Undeclared,” is not just that Biden was declared the winner of the elections. Many other things happened between Nov. 4 and 15, 2020. Trump and many Republican elected officials declared the elections as being rigged, whereas Biden and his team kept saying that the elections were legitimate. The observed results could be due to cognitive dissonance reduction, as the authors claim. But they could also be due to respondents simply being influenced by their party leaders or numerous other things that changed between Nov. 4 and 15. Although the study contains an experimental manipulation (= is a “true experiment”) it is unclear what was manipulated here.

Given the interpretational ambiguities of the results, I feel that the authors overstate their results. I think the authors should adopt more cautious language and delete sentences such as “These findings advance theoretical understanding of polarized perceptions of election legitimacy by more directly implicating rationalizing processes associated with cognitive dissonance and motivated reasoning” and “[…] suggest the role of emotion in the arousal and reduction of dissonance through rationalization.”

Has this study been preregistered? I am asking because the authors made many choices that I would not have made. Here some examples. Party affiliation was measured on a 7-point scale. The authors trichotomized this continuous scale into three categories: Responses 1, 2, and 3 were labeled “Democrat,” response 4 “Independent,” and 4, 6, and 7 “Republican.” Why not a different categorization into responses 1 and 2 (Democrat), 3, 4, and 5 (Independent), and 6 and 7 (Republican)? Or why not treat this variable as a continuous predictor with 7 levels? The latter choice would have made sense because the “quadratic trend” is significant (see Table 1). Where does the idea come from to combine all “other” 14 media outlets into one score which is then contrasted to Fox News?

Given the introduction, I expected the polarizing effect (the increase in difference between Democrats and Republicans between Undeclared and Declared) to be moderated by in-group media consumption. In other words, I expected the authors to test a 3-way interaction between party identification, timing, and media consumption (the latter being a continuous predictor indicating the extent to which respondent consume media known to promote ideas consistent with the respondents’ party identification). I also expected the authors to test a moderated mediation model in which the three-way interactive effect on election legitimacy is moderated by emotions, i.e., the difference between negative and positive emptions. Such a moderated mediation model corresponds to Hayes’ Model #11.

Minor points:

I was confused by the fact that different terms were used interchangeably. “Party identification” was sometimes called “party identity” and “partisan ID.” “Election legitimacy” was sometimes referred to as “vote legitimacy,” “perception of legitimacy,” and “confidence in vote legitimacy.” “Media outlets” are also “news outlets” and “sources.” “Media trust and consumption” is also called “engagement.” “National votes” are sometimes “nationwide votes.”

I didn’t understand the sentence “Respondents answered both questions on non-numeric scales” (p. 13). Are the 5-point scales, which were mentioned two sentences before, non-numeric? A similar issue occurs on page 13 where the authors say “Participants answered on two scales presented without numbers (1 = Not at all confident; 5 = Very confident).” Do the authors mean to say that they presented respondents with five verbal labels and that they later assigned the numbers 1 to 5 to these labels?

The choice of the 15 media outlets is surprising. How come the authors did not include a larger number of conservative outlets?

It is not very informative to compute for each respondent a correlation between trust and consumption across the 15 media outlets, i.e., 1236 correlations each with an N of 15 (p. 13-14). It would be better to compute for each media outlet a correlation across all participants, i.e., 15 correlations each with an N of 1236. The authors should then report the median and the range of these 15 correlations.

On page 14, the authors report a one-df test, F(1, 1078) = 3.13, p = .077, which is a 2-df test.

I’d drop the factor “vote type” in the analyses reported on pages 14-15 and I’d collapse the top and bottom panels in Figure 2). The factor doesn’t add anything to the paper. Given that the two vote confidence ratings are averaged in the remaining analyses, including it as a factor in the earlier analyses creates confusion for readers. The fact that the party identification by timing interaction was stronger for national votes can be mentioned in a footnote.

I did not understand right away what the authors meant by “systematic legitimacy of national votes” (p. 15).

It says on page 15 that “Republicans’ confidence that their own votes were correctly counted did not significantly decrease over time (M_Undeclared= 3.31, […] M_Declared = 3.31)”, but then the red line in the top middle panel in Figure 2 is not perfectly flat. How is this possible?

Figure 2: I’d put the violin density plots for the Independents in the middle rather than on the right side.

The authors say on page 17 “In a regression analysis, negative emotions predicted lower confidence that votes were counted correctly,” but it is not clear what confidence ratings they are referring to, the “own vote,” the “nationwide vote,” or the average of the two ratings. The same issue exists in the title of Table 1 and the analyses reported on p. 20.

It is incorrect to say that Figure 4 is a graphic representation of the result that “Democrats’ and Republicans’ differential engagement with polarized media corresponded with their polarized confidence in vote legitimacy” (p. 19). Confidence in vote legitimacy (= perceived election legitimacy) is not shown in Figure 4.

The significant “curvilinear trend” (b = 0.16) reported in Table 1 is surprising. This coefficient and Figure 2 suggest that the polarizing effect is mostly due to Democrats becoming more confident. Is this finding consistent with the authors’ theoretical analyses? Wouldn’t we expect cognitive dissonance effects be strongest for losers?

I don’t understand why the authors first include media outlet (= source) as a within-subject factor (which is identical to computing a difference score; see p. 18) but then run analyses in which they include both media outlet scores as predictors (Table 1). Is type of media outlet hypothesized to be a moderator, or are the authors predicting the existence of two (parallel?) mediators?

The sentence “It is noteworthy how in essays …” (p. 22) is formulated in an awkward manner.

The authors dedicate four paragraphs to the types of questions that future research might examine (pages 23-25). I think these ideas can be reduced to one paragraph. It would be more interesting for the authors to discuss the implications of their findings rather than provide a list of the numerous things they didn’t do in the present research.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

15 Oct 2021

PONE-D-21-23399

When Election Expectations Fail: Polarized Perceptions of Election Legitimacy Increase with Accumulating Evidence of Election Outcomes and with Polarized Media

Friday, October 15, 2021

Dear Professor Galek:

This letter accompanies the resubmission of PONE-D-21-23399, now titled, “When Election Expectations Fail: Polarized Perceptions of Election Legitimacy Increase with Accumulating Evidence of Election Outcomes and with Polarized Media.” The manuscript is co-authored by me, Marrissa “Dani” Grant, Alexandra Flores, Eric Pedersen, and David Sherman. Note that we have reordered authorship (Grant, Flores, Pedersen, Sherman, & Van Boven) to better reflect the contributions of the two lead authors.

Thank you and the reviewers for the thoughtful and incisive comments. The quality of reviews was of the highest caliber. We have revised the manuscript in response to reviewer comments in ways that, we believe, led to a manuscript with clearer and more substantial contributions. Below, we provide a summary of three larger revisions, followed by detailed responses to your and the reviewers’ comments.

(1) You requested that we conduct the analysis suggested by Reviewer 2, specifically, “to test a 3-way interaction between party identification, timing, and media consumption (the latter being a continuous predictor indicating the extent to which respondent consume media known to promote ideas consistent with the respondents’ party identification).” We now report these analyses in Table 1 and in the Results section (lines 369-382).

The findings are consistent with moderation implied by our hypotheses. To the extent that participants trusted and consumed Fox News, they perceived lower election legitimacy (main effect of Fox News ratings: b = –0.20), that decreased over time (election timing × Fox News ratings: b = –0.17). To the extent participants trusted and consumed other outlets, they perceived higher election legitimacy (main effect of other outlet consumption: b = 0.40) that increased over time (election timing × other outlets ratings: b = 0.28). These findings indicate that trust and consumption of Fox News and other outlets differentially predict perceptions of vote legitimacy.

Moreover, the analysis yielded a 4-way interaction (partisan identification × election timing × other outlets ratings × Fox News ratings: b = –0.29). For Democrats, higher ratings of Fox News was associated with reductions in the increase of perceptions of election legitimacy over time (election timing × Fox News ratings: b = –0.29, t = –3.04, p = .002). Higher Democratic ratings of other outlets were not associated with changes over time in perceived election legitimacy (election timing × Other Outlet ratings: b = –0.04, t = –0.42, p = .676). Among Republicans, in contrast, trust and consumption of other media outlets was associated with a reduced decline in perceived election legitimacy (election timing × Other Outlet ratings: b = 0.54, t = 2.89, p = .040). Higher Republican ratings of Fox News were not associated with changes over time in perceived election legitimacy (election timing × Fox News ratings: b = –0.07, t = –0.74, p = .458). Thus, to the extent that participants trusted and consumed news outlets that are typically distrusted and avoided by their political in-group—Fox News for Democrats and other sources for Republicans—their perceptions of election legitimacy less strongly followed the prevailing pattern within their political group.

As we state in the manuscript, however, we are hesitant to draw overly strong conclusions from an exploratory analysis of higher order interactions. We chose not to explore more complicated moderated mediation and mediated moderation analyses given the potentially complicating presence of a 4-way interaction, the exploratory nature of mediation analyses, and our general hesitancy about unwarranted causal inferences given correlational mediation analyses.

(2) You asked us to scale back reliance on cognitive dissonance theory, depending on the outcome of requested analyses. As noted above, the moderation analysis is generally consistent with hypotheses the the degree to which people trust and consume polarized media sources, they exhibit polarized perceptions of election legitimacy that increase over time. These findings do not allow strong causal conclusions, however. We have adopted more cautious language throughout the paper, clarifying that although we derive predictions from dissonance theory and although our study provides a more precise test of those predictions than previous research on perceptions of election legitimacy, the findings are nevertheless open to alternative interpretations and cannot be exclusively connected to cognitive dissonance theory.

(3) We have revised our theoretical and empirical analyses of emotion. You and the reviewers noted that negative emotion is not a precise measure of dissonance, and that it is difficult to know whether emotions are driving or are driven by polarized perceptions of election legitimacy. We agree and acknowledge that our previous discussion of emotion was overly narrow. Clearly, negative emotions are as much a consequence of perceptions that the election was illegitimate as they are a driver of perceived election legitimacy. Indeed, in an exploratory serial mediation analysis, there was a significant indirect effect from partisan identity to media perception to vote confidence to relative emotion, reflecting that people felt more negative (and less positive) to the extent they were skeptical that votes were counted correctly. Given the reciprocal relationship between emotion and perceived election legitimacy, we report the effects of partisan identity and timing on emotion, and we report the correlations between emotion and perceived election legitimacy (controlling for identity and timing), but we do not make or test causal claims. We have, accordingly, revised our theoretical analysis of emotion in the introduction and elsewhere.

In addition to these three broader revisions, we have made relatively minor revisions in response to comments and suggestions, as detailed below.

EDITOR COMMENTS

Dear Dr. Van Boven,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

I asked two experts in the field to review this manuscript. I synthesize their key comments with my own (based on an independent reading) below. Overall, all three of us think that this paper should be published in PLOS ONE, though we differ somewhat in how much more work is needed.

The two places that I think need to be revised are the positioning and (at least one of) the analyses. For the positioning, you lean heavily on dissonance theory, which certainly makes sense intuitively, and yet as R2 points out, there are other plausible explanations for why belief in the election’s legitimacy could change with time. Beyond that, there is no direct measure of dissonance, making it that much more difficult to know if dissonance is driving this result or not. I would suggest significantly reducing the exposition on dissonance and alluding to the entire idea as just a possibility, rather than a theory to build on (with these results being confirmation of that theory).

See our point (2) above. We have clarified that we derive predictions from cognitive dissonance theory, while acknowledging that our findings may also be consistent with other interpretations. As detailed below, however, part of Reviewer 2’s alternative interpretation, that different media outlets provided different information during the week following election day, is the same explanation implied by cognitive dissonance theory, as articulated in Festinger and colleagues’ “When Prophecy Fails….”

For the analyses, I agree with R2’s belief that the polarization effect should (based on your set up) be moderated by in-group media consumption. I encourage you to run the set of analyses that the reviewer suggested. If the results don't support your theorizing, I would suggest that you both downplay the media consumption angle entirely AND report the lack of a result in an appendix.

As discussed in point (1) above, we conducted the exploratory moderation analysis (lines 369-382).

As discussed in point (3), we acknowledge the reciprocal role of emotion and legitimacy. We report the effects of partisanship and timing on emotion and the correlation between emotion and perceived election legitimacy, controlling for partisanship and timing, but do not report mediation analyses (lines 331-358).

Aside from these two larger issues, I suggest you carefully read and respond to the other comments made by both reviewers as they clearly took the care to provide feedback that would improve your work.

We have addressed the reviewers’ comments, as specified in the following rebuttal letter.

On the whole, I think that with these changes, this manuscript has a clear path to publication. Short of something significantly new in the revision, I do not anticipate sending the manuscript back out to review. As always, please provide a detailed summary of the changes you have made in response to the entire review team’s comments.

Best of luck.

Thank you for the encouragement. We look forward to learning your decision.

REVIEW COMMENTS TO THE AUTHOR

Reviewer #1: I was a reviewer for this manuscript when it was previously submitted to a different journal, and I very much appreciate how the authors have altered the paper to reflect and address many of my earlier concerns. I also thought that the new material from Lee Ross nicely motivated the present investigation. What follows are some lingering issues that I think can easily be addressed in a revision:

Line 82: Unfortunately, I can’t quite parse the following sentence: “For winners, confirming knowledge that their expected and hoped-for win needs to be psychologically reconciled with any lingering doubts about the candidate’s electability and other perceptions of the candidate’s weaknesses” -- even by manipulating whether “confirming” is intended as a verb or an adjective.

Thank you for pointing out the confusing wording. We have revised for clarity (lines 80-82).

Line 102: “The present study also provides evidence for the role of social confirmation in reducing dissonance through motivated reasoning about election legitimacy” – again a bit difficult to parse -- is the role played by social confirmation or motivated reasoning? I realize that “dissonance through motivated reasoning about election legitimacy” is intended to be a single unit, but maybe unpacking the sentence a bit would be helpful.

Thanks for catching this, we have revised this sentence for clarity (lines 96-98).

Line 119: I appreciate the new edit, but “high” should be “highly,” yes?

Yes. We have clarified (line 115).

Line 243: I also appreciate this acknowledgement but maybe it’s also worth acknowledging somewhere that, especially in light of the handy figure, the outcome of the election was somewhat of a continuing process, not a bright-line distinction.

Thank you for pointing out this ambiguity, which we have clarified (lines 195-198).

Line 298-303: Democrats’ conviction that their votes were counted correctly evidently increased much more than the barely significant drop in Republicans’ confidence that their votes were counted correctly, yes? Maybe worth mentioning that difference?

Yes, the increase in Democrats’ confidence that votes were counted correctly was larger in magnitude (MUndeclared = 3.92, SD = 1.00; MDeclared = 4.37, SD = 0.92; F(1, 1201) = 20.98, p < .0001) than the decrease in Republicans’ confidence in vote counts (MUndeclared = 2.99, SD = 1.18; MDeclared = 2.79, SD = 1.24; F(1, 1201) = 3.95, p = .047). We acknowledge this difference (lines 319-323) and speculate that it may simply reflect that Democrats’ increase in vote legitimacy aligned with accumulating evidence that votes were legitimately counted. For Republicans, in contrast, their decreasing confidence in vote counts occurred despite accumulating evidence that votes were correctly counted.

More importantly, why do these means differ from those reported in the next paragraph in Lines 312-313 and Lines 317-318? No doubt I have missed some distinction, but I’d nevertheless appreciate a clarification.

Thank you for pointing out this unclear description. We have chosen to separate these paragraphs into two sections to demonstrate the exploratory nature of the test that includes own vote confidence and nationwide vote confidence separately. In the first section, we report the means and analyses for a perceived election legitimacy which is the average of the own vote confidence and nationwide vote confidence ratings (lines 261-262 for definition of measure, lines 311-318 for results). In an exploratory analysis at the end of the results section, we report separately the means for confidence that own and nationwide votes were correctly counted (lines 431-447). We have also corrected minor errors in reporting means (owing to inadvertent exclusion of people who completed the survey on November 8), which have been corrected.

Line 318-320: “This differential decline may simply reflect that people know more about their own votes than they do about others’ votes.” Perhaps, but then why don’t we see a similar increase for Democrats’ confidence in their own vote vs. others’ votes? Perhaps a ceiling effect?

This is a good point. We have included an acknowledgement that a ceiling effect might have restricted Democrats’ increase in confidence that their own votes were counted correctly (lines 447-449).

Line 437: There appears to be a word missing

We have clarified the sentence (line 491-494).

Line 438: Given the measures employed in the study, it’s a bit odd to refer to, for example, Fox News as a source of collective dissonance reduction in the same way that a doomsday cult (or other tightly-knit social groups) apparently was. The latter obviously offered a much more intimate connection for the target perceiver than the former, which could simply (or primarily) represent a source of biased information, as opposed to all the other ways in which fellow cult members could be a source of comfort when a prophecy is seemingly disconfirmed. Perhaps some such acknowledgment of this distinction is warranted.

We acknowledge the differences more clearly between media outlets and tightly knit social groups (lines 503-507), while speculating that the informational function, if not the social connection functions, of media sources may be similar to social groups.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Reviewer #2: I have to admit that I have never really understood PLOS ONE’s publication criteria and the role of PLOS ONE reviewers. If my task was to examine if the manuscript reports empirical research that satisfies the basic standards of science and if the description of the research is comprehensible for scientists from other fields and practitioners then my answer is “Yes this manuscript should be published in PLOS ONE.” In other words, the manuscript satisfies the 7 criteria listed in the Guidelines for Reviewers.

Below a more nuanced review that highlights the strengths and weaknesses of the manuscript. I am providing rather detailed input to give the authors the opportunity to increase the scientific contribution and impact of their paper.

There are clearly many things to like about this manuscript. The experimental manipulation is a strong point. So is the theoretical research question as well as the attempt to test theoretical predictions about cognitive dissonance in an applied setting. The study is well-conducted, and the analyses are mostly correct and reported in a straightforward manner. I liked the violin density plots. The findings provide evidence for the idea that emotions and perceptions of election legitimacy polarized over time after the 2020 presidential election.

Thank you for the encouraging comments!

Neither the predictions nor the results are particularly surprising. Most people, even those without any training in social sciences, would predict that losers perceive elections as less legitimate (and experience more negative emotions) than winners, and that this difference increases as the outcome of the elections become more and more clear. Once does not need cognitive dissonance to explain this effect.

Thank you for the comments. We do not have evidence that directly addresses whether the results are particularly surprising to social scientists or to the lay public, as we did not collect such data. We would argue, however, that surprisingness or counter intuitiveness are problematic criteria to evaluate scientific contributions. Excessive weighting of counter intuitiveness might even contribute to perverse incentives. A better criterion, we would argue, is whether the findings advance knowledge of how the world is, surprising or not. Most scientific facts, not surprisingly, are not surprising.

The latter part of the comment (“Once [sic] does not need cognitive dissonance to explain this effect”) implies there are alternative theoretical explanations of the findings. We are not sure which specific theoretical explanations the reviewer has in mind, but do not dispute their existence. We have therefore taken care to explain that our predictions are derived from, if not exclusively linked to, cognitive dissonance theory (e.g., lines 17-19, 103-105).

The difference between the two groups, “Declared” and “Undeclared,” is not just that Biden was declared the winner of the elections. Many other things happened between Nov. 4 and 15, 2020. Trump and many Republican elected officials declared the elections as being rigged, whereas Biden and his team kept saying that the elections were legitimate. The observed results could be due to cognitive dissonance reduction, as the authors claim. But they could also be due to respondents simply being influenced by their party leaders or numerous other things that changed between Nov. 4 and 15. Although the study contains an experimental manipulation (= is a “true experiment”) it is unclear what was manipulated here.

Thank you for suggesting this interpretation. This interpretation is broadly consistent with cognitive dissonance theory, and with our suggestion that consumption of polarized media sources contributes to changes in perceived election legitimacy, as discussed in point (2) above. Trump and Biden made different claims about election legitimacy, as did Fox News and NPR. We further suspect that different media sources gave different credence to Trump’s and Biden’s claims of election (il)legitimacy. In both cases, group leaders and political “elites” provided similar dissonance reducing information, with Fox News echoing Trump’s false claims about the election. The moderation results reported in point (1) above implicate the role of consumption of and trust in Fox News versus other sources, but do not strongly differentiate the roles of media versus political leaders. We have acknowledged this overlap in the manuscript (lines 519-521).

Given the interpretational ambiguities of the results, I feel that the authors overstate their results. I think the authors should adopt more cautious language and delete sentences such as “These findings advance theoretical understanding of polarized perceptions of election legitimacy by more directly implicating rationalizing processes associated with cognitive dissonance and motivated reasoning” and “[…] suggest the role of emotion in the arousal and reduction of dissonance through rationalization.”

We have revised the lines you noted and others to use more cautious language (e.g., lines 480-482).

Has this study been preregistered? I am asking because the authors made many choices that I would not have made. Here some examples. Party affiliation was measured on a 7-point scale. The authors trichotomized this continuous scale into three categories: Responses 1, 2, and 3 were labeled “Democrat,” response 4 “Independent,” and 4, 6, and 7 “Republican.” Why not a different categorization into responses 1 and 2 (Democrat), 3, 4, and 5 (Independent), and 6 and 7 (Republican)? Or why not treat this variable as a continuous predictor with 7 levels? The latter choice would have made sense because the “quadratic trend” is significant (see Table 1).

No, the study was not preregistered.

We appreciate the concern about treating party affiliation as a continuous measure. Following established practice in political science and, to a lesser extent, political psychology, we treat party identification as a categorical measure for two reasons. First, the American National Election Study questions are a series of categorical questions. Respondents are first asked whether they identify as Democrat, Republican, or something else. Democrats and Republicans then indicate whether they identify strongly or moderately. Non-identifiers are asked whether they lean Democrat, lean Republican, or do not identify with either party. These responses are fundamentally categorical. Second, and relatedly, the 1-7 scale above lacks interval properties. The change from 1 (strong Democrat) to 2 (moderate Democrat) is arguably different from the change from 4 (Independent) to 5(lean Republican); the latter represents a qualitative shift in group identity.

Where does the idea come from to combine all “other” 14 media outlets into one score which is then contrasted to Fox News?

Contrasting Fox News with other sources reflects research from political science that singles out Fox News as a conservative point of comparison (DellaVigna & Kaplan, 2007) and polling research demonstrating that Republicans trust Fox News more than other sources whereas Democrats distrust Fox News (Gramlich, 2020). We have added these references (lines 173, 221). Our selection of the 15 media sources was guided by the availability of their full text, which we use in a different project (lines 277-281).

Given the introduction, I expected the polarizing effect (the increase in difference between Democrats and Republicans between Undeclared and Declared) to be moderated by in-group media consumption. In other words, I expected the authors to test a 3-way interaction between party identification, timing, and media consumption (the latter being a continuous predictor indicating the extent to which respondent consume media known to promote ideas consistent with the respondents’ party identification). I also expected the authors to test a moderated mediation model in which the three-way interactive effect on election legitimacy is moderated by emotions, i.e., the difference between negative and positive emotions. Such a moderated mediation model corresponds to Hayes’ Model #11.

We address this in point (1) above. We report the results of an analysis with higher order interactions that are consistent with our hypotheses (Table 1, line 388-418). We chose not to conduct mediated moderation or moderated mediation analyses given the complexity of higher order 3-way and 4-way interactions and given the exploratory and tentative nature of such mediation analyses.

Minor points:

I was confused by the fact that different terms were used interchangeably. “Party identification” was sometimes called “party identity” and “partisan ID.” “Election legitimacy” was sometimes referred to as “vote legitimacy,” “perception of legitimacy,” and “confidence in vote legitimacy.” “Media outlets” are also “news outlets” and “sources.” “Media trust and consumption” is also called “engagement.” “National votes” are sometimes “nationwide votes.”

Thank you for pointing out these inconsistencies. We have resolved them.

I didn’t understand the sentence “Respondents answered both questions on non-numeric scales” (p. 13). Are the 5-point scales, which were mentioned two sentences before, non-numeric? A similar issue occurs on page 13 where the authors say “Participants answered on two scales presented without numbers (1 = Not at all confident; 5 = Very confident).” Do the authors mean to say that they presented respondents with five verbal labels and that they later assigned the numbers 1 to 5 to these labels?

We have clarified that respondents did not see numerical labels. We assigned numerical labels for data analyses (lines 258-259, 266-267, 284-285).

The choice of the 15 media outlets is surprising. How come the authors did not include a larger number of conservative outlets?

As noted above, our selection of media outlets was guided by the availability of full text databases for a different project. We acknowledge there are additional conservative media outlets that were not included in the study (lines 277-281, 509-510), while also recognizing that Fox News, as noted above, has been singularly identified in the political science literature as a strong conservative voice among mainstream media (lines 172-175).

It is not very informative to compute for each respondent a correlation between trust and consumption across the 15 media outlets, i.e., 1236 correlations each with an N of 15 (p. 13-14). It would be better to compute for each media outlet a correlation across all participants, i.e., 15 correlations each with an N of 1236. The authors should then report the median and the range of these 15 correlations.

The two correlations address slightly different associations, with convergent evidence bolstering our decision to average ratings of trust and consumption. At the individual level, we compute the correlation between 15 pairs of trust and consumption ratings, which we then average across 1236 respondents (average within-person r = .51). This correlation indicates that, on average, individuals report that they consume more of those media sources that they also rate as more trustworthy, and that this association occurs, on average, across people (lines 285-287). At the level of media outlet, we compute the correlation between pairs of trust and consumption ratings across 1236 respondents, given data availability. We then average those 15 correlations (average between-person r = .51, range: [.36, .67]). This correlation indicates that the more respondents trust a particular media source, the more they report consuming that source, and that this association occurs, on average, across the 15 sources (lines 287-293).

On page 14, the authors report a one-df test, F(1, 1078) = 3.13, p = .077, which is a 2-df test.

This was correctly, but unclearly, reported as a 1-df test of the interaction of timing (Undeclared, Declared) and partisanship (Democrat, Republican), with Independents excluded. We now report the 2-df test of the interaction between partisan identification and timing for clarity (lines 302-303).

I’d drop the factor “vote type” in the analyses reported on pages 14-15 and I’d collapse the top and bottom panels in Figure 2). The factor doesn’t add anything to the paper. Given that the two vote confidence ratings are averaged in the remaining analyses, including it as a factor in the earlier analyses creates confusion for readers. The fact that the party identification by timing interaction was stronger for national votes can be mentioned in a footnote.

Thank you for raising this point. Following your suggestion, we combined the two measures of perceived election legitimacy in the primary analysis (lines 307-323) and in Figure 2, consistent with previous research (Sances & Stewart III, 2015). Our theorizing does not directly imply a difference between these factors.

We disagree, however, that examining the factors separately does not add anything to the paper. Previous research has found systematic differences such that people perceive the health and functionality of democracy at national level to be lower than at the personal level (Sances & Stewart III, 2015). Our study speaks to these differences, so we report analyses at the end of the Results section that includes vote type as a factor (lines 428-447). We speculate in the Results and Discussion that confidence in national vote counting may be more labile in response to motivated reasoning than confidence in own vote counting (lines 443-447, 517-522).

I did not understand right away what the authors meant by “systematic legitimacy of national votes” (p. 15).

For clarity, we have rewritten the introduction to the Perceived Election Legitimacy of Own Vote and Nationwide Vote section to clarify the point we are trying to make with our previous poorly worded statement (lines 428-431).

It says on page 15 that “Republicans’ confidence that their own votes were correctly counted did not significantly decrease over time (MUndeclared = 3.31, […] MDeclared = 3.31)”, but then the red line in the top middle panel in Figure 2 is not perfectly flat. How is this possible?

Thank you for catching this mistake! The correct means are MUndeclared = 3.35 and MDeclared = 3.24, which we corrected in the manuscript.

Figure 2: I’d put the violin density plots for the Independents in the middle rather than on the right side.

Thank you for the suggestion. We have decided to keep the figure in its current state with the order remaining Democrat, Republican, and Independent because our primary comparison of interest is between Democrats and Republicans. In its current order it is easier to visually compare group changes in emotions over election phases when they are placed side-by-side.

The authors say on page 17 “In a regression analysis, negative emotions predicted lower confidence that votes were counted correctly,” but it is not clear what confidence ratings they are referring to, the “own vote,” the “nationwide vote,” or the average of the two ratings. The same issue exists in the title of Table 1 and the analyses reported on p. 20.

We have changed the terminology from “confidence that votes were counted correctly” to “perceived election legitimacy” which is defined as the mean of “nationwide vote confidence” and “own vote confidence” to clarify our analyses (lines 261-266).

It is incorrect to say that Figure 4 is a graphic representation of the result that “Democrats’ and Republicans’ differential engagement with polarized media corresponded with their polarized confidence in vote legitimacy” (p. 19). Confidence in vote legitimacy (= perceived election legitimacy) is not shown in Figure 4.

Thank you for pointing this out. We have clarified that Figure 4 presents media perceptions, without the presentation of perceived election legitimacy (lines 369-370).

The significant “curvilinear trend” (b = 0.16) reported in Table 1 is surprising. This coefficient and Figure 2 suggest that the polarizing effect is mostly due to Democrats becoming more confident. Is this finding consistent with the authors’ theoretical analyses? Wouldn’t we expect cognitive dissonance effects be strongest for losers?

We partially addressed this comment above. It is true that the magnitude of Democrats’ increase in perceived election legitimacy is greater than the magnitude of Republicans’ decrease in perceived election legitimacy. However, interpreting that difference is complicated because the conclusions are not symmetric. For Democrats, consistency motivations and the emerging evidence of correctly counted votes point in the same direction; for Republicans, consistency motivations and the emerging evidence of correctly counted votes point in opposing directions. The combination of these effects imply a larger effect for Democrats than for Republicans. We acknowledge this issue in the Discussion (lines 463-476).

I don’t understand why the authors first include media outlet (= source) as a within-subject factor (which is identical to computing a difference score; see p. 18) but then run analyses in which they include both media outlet scores as predictors (Table 1). Is type of media outlet hypothesized to be a moderator, or are the authors predicting the existence of two (parallel?) mediators?

We analyze ratings of media outlets two ways. First, to capture mean differences between groups, we report a 3(partisan identification: Democrat, Republican, Independent) × 2(media outlet: Fox News, Other Outlets) ANOVA. As you note, the media outlet factor is equivalent to a difference score, and the partisan identification by media outlet interaction reflects the fact that, on average, Republicans and Democrats trust and consume different media sources (Figure 4).

Second, in the moderation analyses (Table 1), we enter ratings of Fox News and other outlets as separate predictors, each interacting with the categorical contrast coded predictors and with each other. This is important because the ratings of Fox News and other outlets are slightly positively correlated at the individual level but negatively related at the mean level. To illustrate, in a series of regressions predicting ratings of Fox News, the effect of ratings of other outlets is not-significant when it is the only predictor (b = 0.01, t(1211) = 0.19, p = .847) but becomes significantly positive (b = 0.33, t(1208) = 7.63, p < .001) when the regression also includes the two contrast codes representing participant partisan identification (Democrat vs. Republican b = 1.28, t(1208) = 15.90, p < .001; Independent vs. Democrat and Republican b = 0.28, t(1208) = 3.09, p = .002). An analogous pattern occurs when predicting ratings of the 14 other sources. In other words, the ratings of Fox News and other sources present something of a Simpson’s paradox (https://en.wikipedia.org/wiki/Simpson%27s_paradox). Given this, it would be inappropriate to analyze ratings of Fox New and Other Outlets as a fixed-effect within person factor in the moderation models.

The sentence “It is noteworthy how in essays …” (p. 22) is formulated in an awkward manner.

We have rephrased the sentence for clarity (line 489-491).

The authors dedicate four paragraphs to the types of questions that future research might examine (pages 23-25). I think these ideas can be reduced to one paragraph. It would be more interesting for the authors to discuss the implications of their findings rather than provide a list of the numerous things they didn’t do in the present research.

We have slightly condensed our suggestions for future research to two paragraphs (lines 517-535). We are hesitant to condense the discussion further because our directions for future research accompany our acknowledgements of shortcomings in the present experiment. We appreciate the suggestion that we discuss the implications of our findings and have expanded the paragraph where we do so (lines 530-535).

References

DellaVigna, S., & Kaplan, E. (2007). The Fox News effect: Media bias and voting. The Quarterly Journal of Economics, 122(3), 1187-1234.

Gramlich, J. (2020, August 18). 5 facts about Fox News. Pew Research Center. https://www.pewresearch.org/fact-tank/2020/04/08/five-facts-about-fox-news/.

Sances, M. W., & Stewart III, C. (2015). Partisanship and confidence in the vote count: Evidence from US national elections since 2000. Electoral Studies, 40, 176-188.

Submitted filename: PONE-D-21-23399-Rebuttal-Letter.docx

Click here for additional data file.

20 Oct 2021

When Election Expectations Fail:

Polarized Perceptions of Election Legitimacy Increase with Accumulating Evidence of Election Outcomes and with Polarized Media

PONE-D-21-23399R1

Dear Dr. Van Boven,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Thank you for the diligent work in responding to the review team's comments/concerns. This was an excellent revision!

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Jeff Galak, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

8 Nov 2021

PONE-D-21-23399R1

When Election Expectations Fail: Polarized Perceptions of Election Legitimacy Increase with Accumulating Evidence of Election Outcomes and with Polarized Media

Dear Dr. Van Boven:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Jeff Galak

Academic Editor

PLOS ONE