This is an open access article distributed under the terms of the Creative Commons
Attribution License (Since the advent of the #MeToo movement, several high-profile politicians have been accused of sexual misconduct. While some members of the public considered these allegations to be legitimate, others have denied or minimised them. This appears to occur even though false sexual harassment allegations are uncommon (Lisak et al., 2010). Although prior work has examined when and why people minimise or deny the legitimacy of sexual misconduct allegations more generally (see Burn, 2019 for a review), little research has examined the factors that contribute to these evaluations in relation to politicians and political figures. Public polling suggests that these patterns of denial and/or minimisation could partially result from partisan bias (e.g. Frankovic, 2020), which is the tendency to evaluate information more favourably when it supports an individual’s existing political beliefs or allegiances than when it opposes them (Ditto et al., 2019; Mason, 2013).
There are several real-world examples that are indicative of a partisan bias in evaluations of sexual harassment allegations. Democratic president Joe Biden was accused of sexually assaulting a former Senate aide, in 1993 (Otterbein & Caputo, 2020), and numerous past sexual assault allegations were made against Republican former president Donald Trump. As a result, the 2020 U.S. Presidential election featured two candidates who had been accused of sexual misconduct, and polling indicated a partisan trend in the perceived credibility of the allegations made against both candidates. According to an Economist/YouGov poll conducted between the 3rd and 5th of May 2020, 14% of Democrats, compared to 55% of Republicans polled, considered the allegations against Joe Biden to be credible (Frankovic, 2020). In contrast, 70% of Democrats and just 14% of Republicans viewed the allegations against then-President Donald Trump as credible. Although we are not suggesting that the allegations, and evidence for or against these allegations, are equivalent in both cases, it appears as though some U.S. citizens make credibility judgements for sexual assault allegations at least in part as a function of their own partisan affiliation.
Against this background, we examined whether political party affiliates react differently to sexual misconduct allegations against male politicians, as a function of the politician’s party affiliation. In other words, are party affiliates more likely to perceive allegations against a politician as more legitimate when the politician is not affiliated with their party? We further tested whether this partisan bias was symmetrical (displayed by both Republicans and Democrats) or asymmetrical (displayed by one party affiliation but not the other). Finally, in a follow-up study, we directly tested a social identity account of partisan bias, by examining whether level of expressive partisanship (i.e., strength of partisan identification) increased partisan bias effects on participants’ perceptions of the legitimacy of these allegations.
Partisan Bias and its Effects
Partisan bias and party identification can work to affect perceptions of the political world for both Democrats and Republicans (Bartels, 2002), and research indicates that partisan-biased information engagement occurs in reaction to a range of different types of scandals against politicians, including corruption (Anduiza et al., 2013), drink-driving, and adultery (Bhatti et al., 2013). Outside of the U.S., an experimental study investigating partisan responses to corruption allegations against politicians from the two major political parties in Spain found that participants who affiliated with the same party as the accused politician rated the allegation as less serious than did participants affiliated with the opposing party to the accused politician. This partisan effect was symmetrical, in that it was consistent for supporters across both political parties (Anduiza et al., 2013). Similarly, in a sample of Danish voters, left-wing voters were less trusting of a right-wing party politician when they were involved in scandals compared to a left-wing party politician involved in scandals, and right-wing voters demonstrated this same partisan bias (Bhatti et al., 2013).
Given these findings, we propose that partisan bias may exist in the context of sexual misconduct allegations against politicians. However, little work to date has experimentally examined whether people engage in partisan bias, as a function of their partisan affiliation, in the context of sexual misconduct allegations against party politicians.
Partisan Social Identification as a Motivator of Partisan Bias
Political and social psychology research suggests that social identification with a political party (i.e. partisan social identity; see Greene, 1999, 2004) can contribute to partisan bias (Huddy & Bankert, 2017; Leeper & Slothuus, 2014). Therefore, should partisan bias exist in responses to sexual misconduct allegations against politicians, partisan social identification may account for it. Partisan social identification research is primarily based on the Social Identity Approach which broadly defines social identity as “those aspects of an individual’s self-image that derive from the social categories to which he [or she] perceives himself [or herself] as belonging” (Tajfel & Turner, 1979, p. 40). That is, social identities are self-definitions that arise from psychologically meaningful group memberships. Social identities can form on the basis of several different types of social categories, including shared political ideologies or parties – such as liberal/Democrat and conservative/Republican (Huddy, 2001; Van Bavel & Pereira, 2018).
Research within the Social Identity Approach suggests that social identification often drives biased reasoning due to individuals’ need for positive distinctiveness and to reduce uncertainty in social contexts. Individuals are therefore motivated to view both their ingroup favourably (and ingroup members as credible and trustworthy) when compared to relevant outgroups (Hornsey, 2008). In the political context, strong partisan identification is likely to lead one to view their political group and its members more positively when compared to individuals in a political outgroup (Huddy, 2001). Individuals who strongly identify with certain social groups often use them to reduce uncertainty in social contexts as group norms can provide an indication of how one should think in circumstances that are ambiguous or unclear (Reicher, 2004). When an individual who has strong partisan identity encounters ambiguous political information, it is likely they will interpret this information in the context of a salient partisan identity to help reduce ambiguity. Given these considerations, examining partisan identification as a social psychological construct allows us to determine the possible role of social identification with a political party in terms of its effect on partisan bias.
The Current Project
In the context of sexual misconduct allegations against party politicians, we propose that partisans would be less likely to view an allegation against a politician of their own party (that is, an aligned or ‘concordant’ politician) as legitimate when compared to the same allegation against a politician from the opposing party (an unaligned or ‘discordant’ politician). Given evidence that partisan biased behaviour can be demonstrated by both Democratic and Republican partisans (Ditto et al., 2019), we expect that both Republican and Democratic affiliates will demonstrate partisan bias in response to such allegations.
Separate from partisan biased responses, it is important to note that, given the nature of the issue, Republicans may be less likely overall to view an allegation of sexual misconduct as legitimate when compared to Democrats. In support of this, Republicans are less likely to consider sexual harassment an important issue (Oliphant, 2017) and are less likely to believe such allegations on average, relative to Democrats (Kulik et al., 2003). Furthermore, conservatives (who are more likely to be Republican) are less concerned about sexual harassment than liberals (van der Linden & Panagopoulos, 2019). Therefore, we would expect a main effect of political affiliation on perceived legitimacy of the allegation, such that Republicans in general may perceive an allegation as less legitimate than Democrats. However, we nonetheless expect that partisans from both political parties will demonstrate partisan bias when evaluating the legitimacy of the allegations.
Understanding the potential role of partisan bias in this domain is important for several reasons. If voters are less likely to engage in objective, non-biased reasoning when determining whether a candidate might have engaged in acts of sexual misconduct, this then could have implications for the likelihood of victims to come forward with such allegations against politicians. Much work has examined factors that influence women’s willingness to label and report sexual harassment and assault, including their endorsement of sexist attitudes (LeMaire et al., 2016), concerns about losing social relationships (Khan et al., 2018), and the extent to which the incident matched their schemas about what assault ‘looks like’ (Peterson & Muehlenhard, 2011). However, no work to our knowledge has focused on whether social identities and the group membership of the accused impacts these tendencies. Thus, it may be that features of the political (and intergroup) context between Republicans and Democrats shape women’s willingness to report sexual harassment owing to their concerns that such allegations may be evaluated in a partisan manner.
Additionally, if there is a political dimension to public judgement of such allegations, this may further influence the likelihood of victim disclosure when the perpetrator is a politician or member of the judiciary, as well as in a high-status position such as being their superior (junior staffer accusing their boss). Therefore, we first explored whether a partisan bias exists in this domain, to experimentally confirm polling that suggests partisan bias in the public’s view of sexual misconduct allegations made against politicians.
The aims of this project were threefold. First, we aimed to determine whether there are partisan differences in reactions to sexual misconduct allegations against politicians. That is, whether Republicans differ from Democrats in their overall responses to sexual misconduct allegations against politicians, irrespective of the party affiliation of the accused politician (tested in Study 1 and replicated in Study 2). Second, we aimed to experimentally test for partisan bias, to determine whether party affiliates judge the legitimacy of such allegations as a function of the party of the accused politician (i.e. more leniently if the politician is of a ‘concordant’ affiliation/less leniently if the politician is of a ‘discordant’ affiliation; tested in Study 1 and replicated in Study 2). Third, we aimed to test a partisan social identity explanation for any partisan bias, by examining whether strength of partisan social identity affects level of partisan bias. That is, whether party affiliates with strong partisan social identity exhibit more partisan bias in their legitimacy judgements of sexual misconduct allegations against politicians (tested in Study 2).
Study 1
Pre-registration for this study can be accessed at https://osf.io/yrhcm. We note and justify deviations from this pre-registration in the Supplementary Materials. Given the above, we hypothesised that:
H1: Republican affiliates would be less likely than Democratic affiliates to perceive a sexual misconduct allegation against any politician as legitimate.
H2: Republicans would be less likely to perceive the allegation as legitimate when made against a Republican candidate (that is, an aligned or ‘concordant’ politician), compared to a Democratic candidate and an unaffiliated candidate (a misaligned or ‘discordant’ politician). A similar effect was predicted for Democrats, such that they would be less likely to perceive allegations against a Democratic politician as legitimate (concordant affiliation) when compared to a Republican or unaffiliated candidate (discordant affiliation).
Method
Design
Study 1 employed a 2 (Participant party affiliation: Democratic, Republican) by 3 (Accused politician’s party affiliation: Democratic, Republican, Unaffiliated) between-subjects design. Perceived legitimacy of the sexual misconduct allegation was the dependent variable.
Participants
Three hundred and ninety-three participants were recruited via Amazon Mechanical Turk. Only workers who indicated that they were affiliates of either the Democratic or Republican parties in their worker profiles were eligible for the study. Fifty-two participants were excluded due to either identifying as an independent in our party screening question, or for failing the pre-registered manipulation check (see ‘Partisan bias manipulation’ subsection below for manipulation check questions). Therefore, the final sample size for this study was 341 (Female = 147, MAge = 34.91, SDAge = 10.96). A majority of Democratic and Republican affiliates were men (see Table 1 for full gender breakdown by political affiliation and experimental condition). All participants received $0.70USD as compensation. A sensitivity analysis for the 2x3 ANOVA conducted in Study 1, for an alpha level of .05 and power set to 80% yields a minimum Cohen’s d of .34. A Cohen’s d effect size of .34 (a small to medium effect size) is equivalent to a partial eta squared of .028.
Materials, Measures and Procedures
Demographic and Political Information
Participants were first asked to report their age, gender, sexual orientation, highest level of education completed, whom they voted for in the 2016 presidential election, political orientation, voting frequency and party affiliation. Political orientation was measured on an 11-pt Likert scale, ranging from 1 (Very Liberal) to 11 (Very Conservative). Party affiliation was measured by asking participants the following question: “In politics today, do you consider yourself a Republican, Democrat, or Independent?”.
Partisan Bias Manipulation
Participants were randomly assigned to read one of three fictitious news stories describing a complaint made against a fictitious male representative (see Supplementary Materials for stimulus material). In each story, the representative (Rep. Doug Williams) was accused of inappropriately touching a female junior staffer. The only variation across the three news stories was the political affiliation of the accused. In the Democratic politician condition, the representative was referred to as “Democrat Doug Williams” in the article title and story, and the party in which the female staffer victim lodged her complaint was referred to as the “Democratic Party”. In the Republican politician condition, “Republican” replaced “Democrat” where “Democrat” was used to refer to the politician and the party in which the victim made her complaint. In the Unaffiliated politician condition, “Politician” replaced political party, and the politician was only referred to as “Rep. Doug Williams”. The political party was referred to as “his political party”, avoiding party labelling altogether. To confirm that participants had engaged with the content in the vignette, participants were required to correctly respond to at least one of three of the following multiple-choice questions: “What was the politician’s gender?”; “What was the junior staffer’s gender?”; and “What political party was the politician representing?”. A further six participants were excluded due to failing this manipulation check by not identifying at least one of these characteristics of the politician or the victim. We report results of the following analysis with the six participants included, for a comparison, in the Supplementary Materials.
Perceived Legitimacy
Six items (α = 0.83, ω = 0.89) measured the perceived legitimacy of the sexual misconduct allegation (e.g., “In your opinion, how accurate is the junior staffer’s account of the incident?”) on a 7-point Likert scale (1 = not at all/not at all likely, 7 = very much/very likely). Higher scores indicated that the complaint was perceived as more legitimate. Though not taken directly from previous research, many studies examining perceptions of sexual misconduct allegations use similar, purpose-built items to assess the perceived legitimacy of such allegations (e.g. Pica et al., 2020). The items used have been included in the Supplementary Materials.
Results
Pre-Registered Analyses
We used R (Version 3.5.1; R Core Team, 2018) and the R-packages apaTables (Version 2.0.4; Stanley, 2018), car (Fox & Weisberg, 2019), lavaan (Rosseel, 2012), dplyr (Version 1.0.2; Wickham et al., 2020), lm.beta (Behrendt, 2014), psych (Version 2.0.9; Revelle, 2020), readr (Version 1.4.0; Wickham & Hester, 2020), and tableone (Version 0.12.0; Yoshida & Bartel, 2020) for all of our analyses.
Descriptive Statistics
Table 1 (below) provides the gender breakdown by both political affiliation and experimental condition, as well as mean scores and standard deviations for Perceived Legitimacy across party affiliation and experimental condition.
Table 1
Descriptive Statistics, Stratified by Condition, Provided for All Participants, and Separately for Democratic and Republican Affiliates, With Gender Breakdown Included
| Variables | Total | Democratic Pol. | Republican Pol. | Unaffiliated Pol. |
|---|---|---|---|---|
| All participants | N = 341 | N = 114 | N = 117 | N = 110 |
| Gender (Male) | 194 (57%) | 67 (58.8%) | 70 (59.8%) | 57 (51.8%) |
| Perceived Legitimacy | 5.23 (1.06) | 5.30 (.94) | 5.28 (1.10) | 5.12 (1.14) |
| Democratic Affiliates | N = 182 | N = 63 | N = 61 | N = 58 |
| Gender (Male) | 96 (52.7%) | 39 (61.9%) | 33 (54.1%) | 24 (41.4%) |
| Perceived Legitimacy | 5.47 (1.11) | 5.37 (1.03) | 5.68 (1.05) | 5.35 (1.23) |
| Republican Affiliates | N = 159 | N = 51 | N = 56 | N = 52 |
| Gender (Male) | 98 (61.6%) | 28 (54.9%) | 37 (66.1%) | 33 (63.5%) |
| Perceived Legitimacy | 4.96 (.94) | 5.21 (.82) | 4.84 (.98) | 4.86 (.99) |
Note. Means, and standard deviations (in parentheses), provided for all continuous variables. No participants identified as gender non-binary.
Perceived Legitimacy
As shown in Table 2, there was a significant main effect for Party Affiliation. In line with H1, Republican affiliates were less likely to evaluate sexual harassment allegations as legitimate (M = 4.96, SD = .94), than Democratic affiliates (M = 5.47, SD = 1.11), regardless of the accused politician’s party affiliation. There was no main effect of condition, indicating that the perceived legitimacy of the complaint did not differ when the politician was affiliated with the same party as the participant compared to when the politician had an ambiguous or opposing affiliation to the participant. Additionally, the interaction of party affiliation and condition was significant.
Table 2
2x3 ANOVA Results Using Perceived Legitimacy of Allegation as the Criterion (N = 341)
| Predictor | Sum of Squares | df | Mean Square | F | p | |
|---|---|---|---|---|---|---|
| Affiliation | 21.6 | 1 | 21.6 | 20.42 | < .001 | .06 |
| Condition | 2.0 | 2 | 1.00 | .95 | .389 | .01 |
| Affiliation x Condition | 6.56 | 2 | 3.28 | 3.10 | .046 | .02 |
| Residuals | 354.53 | 335 | 1.06 |
However, contrary to H2, Tukey’s HSD post-hoc testing demonstrated that there were no significant mean differences in perceived legitimacy for Republican affiliates in the Republican politician condition (M = 4.84, SD = .98), compared to Republican affiliates in the Democratic politician (M = 5.21, SD = .82; p = .450) and Unaffiliated politician (M = 4.86, SD = .99; p = .999) conditions, respectively. Similarly, there were no mean differences between Democratic affiliates in the Democratic politician condition (M = 5.37, SD = 1.03) compared to the Republican politician (M = 5.68, SD = 1.05; p = .538) and Unaffiliated politician (M = 5.35, SD = 1.23; p = .999) conditions.
Discussion
The first aim of Study 1 was to determine whether partisan differences exist in reactions to sexual misconduct allegations against politicians, in general. The second aim was to experimentally test whether partisan bias exists in judging the legitimacy of a sexual misconduct allegation against a male politician. Broadly consistent with the literature (van der Linden & Panagopoulos, 2019) and opinion polling (Oliphant, 2017), and in support of H1, self-identified Republican affiliates were less likely than self-identified Democratic affiliates to perceive any sexual misconduct allegation as legitimate. This suggests that a woman’s sexual misconduct allegation against a male politician, irrespective of the political party of the accused, is more likely to be accepted and believed by Democrats than Republicans. However, H2, that participants would find allegations against a politician of discordant affiliation more legitimate when compared to allegations against a politician of concordant affiliation was not supported. On average, Republican affiliates in the Republican politician condition did not judge the legitimacy of the allegations differently to Republican affiliates in either the Democratic or Unaffiliated politician conditions. A similar pattern of results was obtained for Democratic affiliates. These results suggest that party affiliates do not judge the legitimacy of sexual misconduct allegations against politicians in a manner that is biased in favour of their partisan affiliation, and their legitimacy judgements may instead be driven by underlying differences between Republicans and Democrats that are common within each party. That is, Republicans are simply less likely to perceive a sexual misconduct allegation as credible, relative to Democrats, independent of the party affiliation of the accused politician. Interestingly, participants appeared to judge the legitimacy of allegations similarly for an unaffiliated politician and a concordant politician, which is consistent with Anduiza et al. (2013), who found that participants judged the seriousness of corruption allegations against a same party politician as similar to a politician of unknown party affiliation.
Study 2
Study 1 was limited in two important ways. First, a sensitivity analysis conducted after completion of the study yielded an effect size equivalent to a partial eta squared of .028, which is larger than the effect size obtained for our interaction effect ( = .02). Therefore, a larger sample size is required to detect smaller partisan bias effects.
Second, Study 1 was limited by its treatment of partisan affiliation as a categorical variable. That is, strength of partisan identification was not measured. It is possible that partisan bias exists at higher levels of partisan identification, as the stronger one identifies with their respective political party, the more likely they are to perceive the allegation according to their political party’s norms and beliefs, and therefore attempt to bolster the positive self-image of the group (Ellemers et al., 1997). We test this possibility in Study 2 by examining the moderating role of expressive partisanship on the effect of affiliation discordance on perceived legitimacy of the sexual misconduct allegation.
Finally, there were no participant party affiliation differences on perceived legitimacy of allegations between participants in the unaffiliated politician condition and the affiliated (Democratic or Republican) politician conditions, respectively. Therefore, we decided to only compare responses to allegations against Democratic and Republican politicians, omitting the unaffiliated condition. In addition, given that most politicians are party representatives as opposed to identifying as independent or unaffiliated, particularly in the U.S. political context, party affiliates will usually be aware that an accused politician is a political ingroup or outgroup member. As such, the primary focus of examining partisan bias in this domain ought to be on politicians with a party affiliation.
Expressive Partisanship
Expressive partisanship is a type of partisan social identification which refers to an affective attachment and sense of belonging to a political party (Huddy & Bankert, 2017). As such, people with high levels of expressive partisanship are motivated to defend their party's status (Bankert et al., 2017). Recent research suggests that expressive partisans are more likely to engage in motivated reasoning due to this need to improve party standing (Leeper & Slothuus, 2014). Therefore, they might be more likely to defend a candidate or representative from their own party against sexual misconduct allegations, and/or be more likely to believe that a representative from the other party has engaged in wrongdoing. Thus, Study 2 examined whether strength of expressive partisanship moderated partisan bias (i.e. discordance) on the perceived legitimacy of the allegation. Study 2 also acted as a replication of Study 1, in that we also tested for an overall difference in responses to sexual misconduct allegations between party affiliations, as well as a partisan bias effect. Pre-registration for this study can be accessed at https://osf.io/mx9jt. Once again, explanations for the deviations from pre-registration can be found in Supplementary Materials.
Given the above, we hypothesised that:
H1a: Compared to Democratic affiliates, Republican affiliates will perceive sexual misconduct allegations against a politician of any party as less legitimate.
H1b: Individuals who rate the legitimacy of an allegation against a politician with a discordant affiliation (i.e., politician affiliation differs to participant political affiliation) will rate the allegation as more legitimate than individuals who rate the legitimacy of an allegation against a politician with a concordant affiliation (i.e., politician affiliation is the same as participant affiliation).
H1c: The negative effect of affiliation discordance on perceived legitimacy will be stronger for individuals higher in expressive partisanship. Participants with higher expressive partisanship will perceive the legitimacy of sexual misconduct allegations as more legitimate when the accused has a discordant affiliation, compared to participants with lower expressive partisanship.
Method
Design
Study 2 employed a 2 (Participant party affiliation: Democratic, Republican) by 2 (Accused politician’s party affiliation: Democratic, Republican) between-subjects design, with expressive partisanship as a continuous measured moderator. Perceived legitimacy of the sexual misconduct allegation was the dependent variable.
Participants
Six hundred and seventy-two participants were recruited via Amazon Mechanical Turk. As in Study 1, only workers who noted that they were affiliates of either the Democratic or Republican parties in their worker profiles were eligible to participate. Sixty-eight people who identified as political independents were excluded from the analysis. A further two were excluded for failing preregistered manipulation checks. The same three check questions used in Study 1 were employed in Study 2. However, the inclusion criteria were more stringent than in Study 1. Participants had to correctly identify the political party of the accused, the gender of the accused politician, and the gender of the staffer, whereas in Study 1 they only had to correctly identify one of the three facts within the news story. Details of party affiliation screening, as well as the same analyses conducted as below with the two excluded participants included, are in the Supplementary Materials.
The final sample size was 602 (Female = 304, non-binary = 1, not disclosed = 2, Mage = 40.61, SDage = 12.66). There was an even split between Democratic and Republican affiliates (301 each). The majority of Democratic affiliates were women, although there was a roughly even split across gender for Republican affiliates, with slightly more men than women (see Table 3). All participants received $0.70USD as compensation. A sensitivity analysis conducted for the moderated linear regression, for an alpha level of .05, a sample size of 602, and power set to 80% yields a minimum detectable R2 of .020, which is conventionally a small effect.
Materials, Measures and Procedure
Materials, the outcome measure of perceived legitimacy (α = 0.84, ω = 0.90) and the procedure for this study remained the same as Study 1 with two differences. The unaffiliated politician condition was not used in this study, therefore there were only two conditions: Republican male politician accused of sexual misconduct and a Democratic male politician accused of sexual misconduct, both again referring to a fictitious representative Doug Williams. We also added the Expressive Partisanship scale (below) to measure participants’ strength of identification with their party.
Expressive Partisanship
Expressive Partisanship was measured using Bankert, Huddy, and Rosema’s (2017) Partisan Identity Scale (α = 91; ω = .93; for both Democratic and Republican versions of the scale). This is an eight-item, seven-point scale (1 = Strongly Disagree, 7 = Strongly Agree) that measures strength of partisan identification with either the Republican Party or the Democratic Party. Party affiliates only received the scale relevant to their stated party affiliation and did not receive the opposing party’s partisanship scale. This scale was administered prior to the experimental manipulation, with higher scores indicating stronger expressive partisanship. An example item is: “When I speak about the Republican/Democratic Party, I usually say ‘we’ instead of ‘they.’”
Results
Pre-Registered Analyses
We conducted a hierarchical moderated linear regression. In Step 1, the direct effects of Party Affiliation (coded as Democratic affiliation = 1, Republican affiliation = 2), Affiliation Discordance (coded as Concordance = 1, Discordance = 2), and Expressive Partisanship were included to test for main effects of Party Affiliation (H1a), Discordance (H1b) and Expressive Partisanship. Two interaction terms, Affiliation Discordance*Party Affiliation and Affiliation Discordance*Expressive Partisanship (H1c) were added at Step 2 to examine whether any Discordance or Expressive Partisanship effects were conditional on party affiliation. We used the same statistical software (R version 3.5.1) and the same R packages to conduct all analyses as in Study 1.
Affiliation Discordance
Affiliation discordance is a dichotomous variable derived from both the participant’s political affiliation and the experimental condition they were assigned to. Participants were classified as concordant if the accused politician’s party affiliation in the experimental condition to which they were randomly allocated to matched their own (e.g. a participant with a Democratic affiliation in the Democratic politician condition). If a participant was allocated to an experimental condition with a politician’s affiliation that was different to their own (e.g. a participant with a Democratic affiliation in the Republican politician condition) they were classified as discordant. This allowed us to compare perceived legitimacy for participants who viewed a news story of a politician in their own party compared to those who viewed a news story of a politician from the opposing party. Put simply, a significant main effect of affiliation discordance (such that discordant participants found the allegations in the fictitious news story significantly more legitimate than concordant participants) would indicate a partisan bias effect.
Coding our participants in this manner allowed us to test an overall partisan bias effect more easily. This is because this coded variable allowed us to directly compare perceived legitimacy between participants who were allocated to a condition where the accused politician in the vignette was aligned to their (the participant’s) political party affiliation versus those participants who were allocated to an experimental condition where the politician in the vignette was not aligned with their affiliation. Furthermore, the Discordance*Party Affiliation interaction term allowed us to test whether the bias was symmetrical (displayed by both Republican and Democratic affiliates) or asymmetrical (displayed by one group of affiliates but not the other). A nonsignificant effect for this interaction term would indicate that any partisan bias effect found for the direct effect of Discordance was not a result of one specific party affiliation driving this direct effect by engaging in more partisan bias than the other.
Descriptive Statistics
Table 3 (below) provides the gender breakdown by both political affiliation and experimental condition, as well as mean scores and standard deviations for Perceived Legitimacy and Expressive Partisanship across party affiliation and experimental condition.
Table 3
Descriptive Statistics, Stratified by Condition, Provided for All Participants, and Separately for Democratic and Republican Affiliates, With Gender Breakdown Included
| Variables | Total | Democratic Pol. | Republican Pol. |
|---|---|---|---|
| All Participants | N = 602 | N = 303 | N = 299 |
| Gender (Male) | 295 (49.2%) | 138 (45.5%) | 157 (52.9%) |
| Total Expressive Partisanship | 4.34 (1.22) | 4.39 (1.25) | 4.29 (1.17) |
| Perceived Legitimacy | 5.27 (.94) | 5.36 (.87) | 5.19 (1.00) |
| Democratic Affiliates | N = 301 | N = 151 | N = 150 |
| Gender (Male) | 140 (46.8%) | 65 (43.0%) | 75 (50.7%) |
| Expressive Partisanship | 4.34 (1.22) | 4.39 (1.26) | 4.28 (1.19) |
| Perceived Legitimacy | 5.60 (.90) | 5.58 (.90) | 5.61 (.90) |
| Republican Affiliates | N = 301 | N = 152 | N = 149 |
| Gender (Male) | 155 (51.5%) | 73 (48.0%) | 82 (55.0%) |
| Expressive Partisanship | 4.34 (1.21) | 4.40 (1.25) | 4.29 (1.17) |
| Perceived Legitimacy | 4.95 (.87) | 5.14 (.77) | 4.76 (.92) |
Note. Means, and standard deviations (in parentheses), provided for all continuous variables. Three participants identified as gender non-binary.
Perceived Legitimacy
As shown in Table 4 below, and consistent with Study 1, there was a main effect of party affiliation. Republican affiliates (M = 4.95, SD = .87) were significantly less likely than Democratic affiliates (M = 5.60, SD = .90) to perceive the allegation as legitimate, regardless of the political party of the accused. However, unlike Study 1, participants with a discordant affiliation to the accused politician were more likely to view the allegation as more legitimate than participants with a concordant affiliation to the accused, which indicates a partisan bias effect. Additionally, the effect of discordance on perceived legitimacy was moderated by party affiliation in Step 2, suggesting that the partisan bias effect may be a function of which party a person is affiliated with. However, expressive partisanship did not moderate the effect of discordance on perceived legitimacy of the allegation, suggesting that strength of partisan identification does not increase partisan bias in this domain.
Table 4
Hierarchical Moderated Multiple Regression Predicting Perceived Legitimacy of Allegation (N = 602)
| Variable | B | 95% CI | β | SE | p |
|---|---|---|---|---|---|
| Step 1a | |||||
| Party Affiliation (Republican) | -.65 | [-.79, -.51] | -.35 | .07 | < .001 |
| Discordance | .20 | [.06, .34] | .11 | .07 | .006 |
| Expressive Partisanship | .03 | [-.03, .09] | .04 | .03 | .305 |
| Step 2b | |||||
| Party Affiliation (Republican) | -.82 | [-1.02, -.62] | -.44 | .10 | < .001 |
| Discordance | .17 | [-.38, .71] | .09 | .28 | .545 |
| Expressive Partisanship | .05 | [-.04, .13] | .06 | .04 | .281 |
| Discordance x Affiliation | .35 | [.07, .63] | .16 | .14 | .016 |
| Discordance x Expressive Partisanship | -.03 | [-.15, .08] | -.08 | .06 | .579 |
aR2 = .13. bR2 = .14; ΔR2 = .01.
To examine the nature of the discordance*party affiliation interaction, we conducted two separate linear regressions for Republican and Democratic affiliates respectively (see Table 5).
Table 5
Linear Regressions for Perceived Legitimacy of Allegation for Democratic Affiliates and Republican Affiliates, by Affiliation Discordance (N = 602)
| Party affiliation | Variable | B | 95% CI | β | SE | p |
|---|---|---|---|---|---|---|
| Democratic Affiliates (N = 301) | Discordance | .03 | [-.17, .23] | .02 | .10 | .777 |
| Republican Affiliates (N = 301) | Discordance | .37 | [.18, .57] | .22 | .10 | < .001 |
For Republicans, affiliation discordance was a significant positive predictor of perceived legitimacy, such that Republican affiliates perceived the allegations against the Democratic politician as more legitimate (M = 5.14, SD = .77) than the allegations against the Republican politician (M = 4.76, SD = .92). However, there was no significant effect of discordance for Democrats. Figure 1 displays the distributions of legitimacy scores for Democratic and Republican affiliates separately, for each condition.
Figure 1
Jitter Bar Plot Comparing the Distribution of Scores on Perceived Legitimacy Across Party of Accused Politician Conditions, for Each Group of Party Affiliates
Note. Horizontal bars represent group means with bootstrapped 95% confidence limits.
Discussion
The aim of the second study was to test a partisan social identity account of partisan bias in perceived legitimacy of sexual misconduct allegations against party politicians, as well as to replicate the test of a general partisan difference effect found in Study 1, and to again test a partisan bias effect that was not found in Study 1. In line with Hypothesis 1a, Republican affiliates evaluated sexual misconduct allegations overall as less legitimate than Democratic affiliates. This provides further evidence that Republicans, irrespective of the political party of the accused politician, are less likely than Democrats to perceive sexual misconduct allegations against politicians as legitimate. In line with Hypothesis 1b, although not found in Study 1, participants who read the allegation against a politician with a discordant party affiliation to their own were more likely to perceive the allegation as legitimate than participants who read the allegation against a politician with a concordant affiliation. However, this effect was asymmetrical. Republicans, but not Democrats, were significantly more likely to perceive allegations against a discordant candidate as more legitimate than a concordant candidate.
Finally, Hypothesis 1c, that the negative effect of affiliation concordance on perceived legitimacy would be stronger for individuals higher in expressive partisanship, was not supported. Higher levels of participants’ social identification with their political party did not increase their tendency to perceive an allegation against a discordant politician as more legitimate compared to a concordant politician. This suggests that this partisan effect is not reliant on how strong one identifies with their respective political ingroup, and that instead simply being a member of said political group is enough to produce these outcomes (for Republicans at least).
General Discussion
The aims of this research were threefold. First, we sought to test whether there was a difference in legitimacy perceptions of sexual misconduct allegations between Democratic and Republican affiliates (Studies 1 and 2). Second, we aimed to experimentally confirm partisan bias observed in opinion polling statistics suggesting that some party affiliates judge the legitimacy of sexual misconduct allegations according to the political party of the accused (Studies 1 and 2). Third, we tested a partisan social identity account of such an effect, that partisan bias in perceived legitimacy of such allegations might be conditional on level of expressive partisanship (Study 2). We were also able to examine whether any partisan bias was symmetrical (occurring in affiliates from both parties) or asymmetrical (occurring only for Democratic or Republican party affiliates).
Overall, findings from both studies suggest that Republicans are less likely than Democrats to perceive sexual misconduct allegations as legitimate, which is broadly consistent with research demonstrating that liberals are more likely than conservatives to be concerned with sexual misconduct (van der Linden & Panagopoulos, 2019). Findings from Study 2 (but not Study 1) demonstrated that party affiliates may judge these allegations in a biased manner when levelled against male party politicians. In Study 2, participants with a discordant affiliation to the accused politician in their allocated vignette (i.e. a politician from the opposing party to the participant) were generally more likely to perceive the allegations as legitimate than participants with a concordant affiliation with the politician (i.e. a politician from the same party as the participant). Furthermore, this partisan bias may be asymmetrical. Republicans, but not Democrats, were more likely to perceive an allegation against the opposing (discordant) politician as more legitimate than the same allegation against the same party (concordant) politician. Finally, in testing a partisan social identity account of partisan bias, we found that expressive partisanship did not influence the discordance effect. The implications of these findings are discussed below.
Partisan Asymmetry of Perceived Legitimacy and Partisan Bias
While Republican affiliates were less likely to perceive sexual misconduct allegations as legitimate, regardless of the political party of the accused, Study 2 showed that they were also less likely to perceive an allegation as legitimate when made against a Republican politician compared to a Democratic politician. This may indicate that Republicans are more sceptical of sexual misconduct allegations in general, compared to Democrats, due to key ideological and moral value differences. Conservatives (who tend to be Republican affiliates) appear less concerned about sexual misconduct in general (van der Linden & Panagopoulos, 2019), indicating a potential ideological difference between Republicans and Democrats on the issue. However, this only explains general party affiliation differences on sexual misconduct, and not the asymmetry of partisan bias witnessed in Study 2. For this, we propose two possibilities to consider not only in the context of this issue but more broadly in other domains where partisan bias in judgement exists.
Republicans might be displaying a heightened moral concern for ingroup loyalty over concern for the harm of others (Rempala et al., 2016). Research adopting a Moral Foundations Theory approach suggests that conservatives, compared to liberals, have a stronger preference for the moral foundation of ingroup loyalty (Koleva et al., 2012). In the context of our study, Republicans may be displaying such loyalty to their politician in the face of accusations, potentially affording them more benefit of the doubt than a Democratic politician. In an aforementioned study, conservatives were more likely to condemn an outgroup member (Harvey Weinstein – Hollywood “liberal”) compared to an ingroup member (Bill O’Reilly – conservative commentator) in the context of sexual misconduct allegations (van der Linden & Panagopoulos, 2019). Whereas for Democrats, harm and fairness are stronger moral considerations, relative to the moral foundation of loyalty. Therefore, they are more willing on average to consider the possibility that an ingroup member has caused someone harm. On issues where an individual or group is being harmed by an ingroup member, Democrats may be more likely to be concerned with harm over ingroup loyalty, and therefore not engage in a partisan manner.
Another potential explanation for the asymmetry in partisan bias found in Study 2 derives from research examining the role of subordinate identities in partisan politics. Democratic and Republican affiliates have social identities that may relate to, and indeed result in, formation of their partisan identification (Klar, 2013; Mason, 2018). This research indicates that the Democratic and Republican higher order social categories are each made up of a multitude of lower order social groups and identities. For example, Democrats may identify as feminists, specific ethnic minorities, and/or young people, whereas Republicans may identify as evangelical and/or European-American. These subordinate identities may play a role in the level of partisan bias displayed on a given issue, particularly when those subordinate identities are salient and are activated (Klar, 2013). In the context of sexual misconduct allegations, a Democrat with a feminist identification may experience competing social psychological motivations, and ultimately behave in a manner more consistent with their feminist identification over their party identification, and therefore denounce their accused politician (Mason, 2018). Psychological needs arising from these subordinate identities may override the need to defend one’s partisan affiliate group. Also, in this context, Republicans may be less likely to experience these competing demands, and instead fall back to engaging in partisan bias due to a lack of pressure on their subordinate identities.
No Effect of Strength of Partisan Social Identity
Findings from Study 2 suggest that strength of partisan social identification (operationalised as expressive partisanship) does not influence partisan bias in this domain. One possible explanation is that a positive partisan identification may not affect perceptions in this domain as much as negative opinions of the outgroup political party and its members. Research on affective polarisation (that is, divergence in feelings between ingroup and outgroup parties, such that outgroup partisans are perceived more negatively and ingroup partisans are perceived more positively), can drive perceptions of political opponents as illegitimate (Iyengar & Westwood, 2015). From this perspective, party affiliates are not affording leniency towards concordant party politicians but are instead judging discordant party politicians more severely. In both studies, legitimacy judgements of the allegations were, on average, higher than the midpoint of the perceived legitimacy scale in both studies. This may indicate that people find all allegations relatively legitimate, but allegations against opposing party politicians, relative to same party politicians, are particularly accurate and truthful. Additionally, mean legitimacy scores in Study 1 in the unaffiliated politician condition were similar to mean legitimacy scores for concordant politicians, suggesting that differences in legitimacy allegations may be a function of harsher judgements of opposing party politicians rather than leniency towards ingroup party politicians.
Limitations and Future Directions
This research provided an experimental test of partisan bias in the domain of sexual misconduct allegations against male politicians in the United States. As such, these findings may be specific to the US political context, and conclusions may not be generalised to other political contexts in other nations. Furthermore, these findings cannot be generalised to other potentially ambiguous informational contexts involving party politicians, such as other types of scandals.
Several factors that could have influenced our results were not controlled for, such as the relative severity of the allegation or the weight of independent evidence afforded to the allegation. Finally, although this research provides experimental support for a partisan bias effect in perceived legitimacy of sexual misconduct allegations, we are unable to confidently point to underlying explanatory mechanisms. Future research should look to examine the asymmetry in partisan bias in this domain, by examining differences in ingroup loyalty across partisans. It should also control for the effects of ideological beliefs such as feminism, and subordinate identifications such as female and feminist identification among Democrats.
Conclusion
Across both studies, our results show that on average, Democratic affiliates perceive sexual misconduct allegations against politicians as more legitimate than do Republican affiliates. Although no partisan bias effects were found in Study 1, a follow-up replication study with a larger sample found that Republicans, but not Democrats, are more likely to judge the legitimacy of sexual misconduct allegations against politicians in a partisan-biased manner. Republicans appear to view allegations against a Democratic politician as more legitimate than allegations against a Republican politician. Unexpectedly, partisan bias was not conditional on level of expressive partisanship, indicating that mere affiliation, rather than strong identification, with the Republican Party is enough for some Republicans to perceive allegations as more legitimate if they are against a Democratic politician than a Republican politician. A result of this polarised political environment may be that victims of sexual misconduct by party politicians feel less comfortable raising the allegations in the public domain, particularly if the politician is a Republican. Partisan bias should be examined not just as a function of partisan identity, but core similarities and differences in ideological beliefs, moral values, subordinate identifications, and negative feelings towards discordant party members, which could all be pertinent in a context which gives rise to partisan bias.