This is an open access article distributed under the terms of the Creative Commons
Attribution License (Understanding the effects of politics on moral judgment requires investigating multiple sources of political influence. The moral judge herself is not the only bearer of political motivations; the source where she learns of morally relevant events, such as a media outlet, may have its own political slant. Furthermore, the individual accused of wrongdoing may also have a recognizable political orientation. The present research isolates three different aspects of political influence in four studies focused on moral judgments concerning sexual misconduct: (1) the influence of the political orientation of media outlets; (2) the influence of moral judges’ political orientation; and (3) the influence of alleged perpetrators’ political orientation. We examine how these factors influence moral judgments targeting both the accusers (alleged victims) and the accused (alleged perpetrators). We demonstrate that attributions of causal contribution, responsibility, and blame are politically influenced, for both actual news reports and fictional scenarios about sexual misconduct.
These studies contribute to closing a gap in the social and moral psychology literature concerning the role of politics in causal and moral judgment of sexual violations. In addition, this work illuminates responses to the global #MeToo movement against sexual harassment, sexual assault, and sexual misconduct (New York Times, Bennett, 2018; Time Magazine, Dockterman, 2018), which has facilitated numerous highly-publicized reports of sexual misconduct. Reporting on the movement has involved alleged offenders with well-known political leanings; however, it is not known whether and how politics systematically influence moral judgment of sexual misconduct cases. For example, while most Americans believe that sexual harassment is a “somewhat-to-very serious” problem, there appear to be partisan divides: Republicans, compared to Democrats, are less likely to view sexual misconduct as a societal problem (Panagopoulos & van der Linden, 2018). Scientific study is needed to clarify whether partisan divides are evident in specifically moral judgments concerning sexual misconduct.
The Influence of Media Politics
There has been considerable debate over whether different news outlets report stories in a way that is biased toward either end of the political spectrum. Complaints about liberal bias in the news are ubiquitous in the United States, drawing the attention of both scholars (e.g., Lee, 2010) and pundits (Siegfried, 2018). At the same time, those with liberal views repeatedly voice concern about media bias, pointing to the broad reach of conservative-leaning news conglomerates (Martin & McCrain, 2019). Social scientists have investigated the existence and prevalence of the alleged political biases in the media, and whether those biases influence media audiences’ beliefs and behaviors. On the one hand, some research indicates that media outlets are regularly partial to particular political parties or figures (Gentzkow & Shapiro, 2010; Martin & Yurukoglu, 2014), and this may contribute to media consumers favoring certain political parties and election candidates (Druckman et al., 2018; Druckman & Parkin, 2005; Kahn & Kenney, 2002). On the other hand, additional studies suggest that while partisan bias may be common in reporting on political scandals, there is little evidence of partisan bias more broadly (e.g., Budak et al., 2016; D’Alessio & Allen, 2000; Niven, 2003). Thus, while some evidence points to an effect of media bias on media consumers, there is limited reason for confidence that the effect obtains across all reporting domains, and little work directly addresses sexual misconduct.
Political Orientation and Moral Judgment
In addition to research on potential political bias in the media, the role of the politics of news consumers has received scientific attention. Research suggests that both conservative and liberal individuals tend to seek out news outlets that accord with their own political views (Iyengar & Hahn, 2009; Knobloch-Westerwick & Meng, 2011; Knobloch-Westerwick et al., 2020). Indeed, it has been proposed that any effect of media bias could largely be consumer-generated: even though online news consumption entails more exposure to opponents’ views, people may create “echo chambers,” where they repeatedly return to a few favored news outlets for information (Flaxman et al., 2016).
There is a pervasive psychological tendency for people to seek out and more readily accept information produced by news media when they align with their own priors (Lodge & Taber, 2013; Stanley, Henne, et al., 2020; Washburn & Skitka, 2018). Across a variety of cultural and political issues, people with conservative and liberal views actively avoid exposure to reasons, arguments, and evidence for opposing views, because they anticipate that such exposure will be frustrating to them, or even damage their relationships with intimates expressing such positions (Frimer et al., 2017; Stanley, Whitehead, et al., 2020; however, see Dorison et al., 2019). Rejection of opponents’ ideas and preferential treatment of one’s ingroup is also supported by mechanisms such as confirmation bias (Nickerson, 1998) and group polarization (Sunstein, 2002), where individuals interpret information so that it aligns with their preexisting beliefs, and double-down on judgments that reflect and promote their ingroup.
In addition, empirical research indicates that people’s politics influence their judgments about the victims and perpetrators of sexual assault (e.g., Lambert & Raichle, 2000). One pathway for this effect might involve individual differences in moral values. Research on Moral Foundations Theory (Graham et al., 2011) has shown that people vary in the extent to which they endorse (at least) five different moral values – care, fairness, loyalty, authority and purity – as a function of political orientation (Graham, Haidt, & Nosek, 2009). In many cultures across the globe, people with more conservative views tend to endorse all five values, whereas people with more liberal views tend to value care and fairness over the “binding” values of loyalty, authority and purity (Doğruyol et al., 2019; Graham et al., 2011; Hu et al., 2020). Roughly, more liberal individuals emphasize the importance of protecting individuals from harm and unfairness regardless of group membership, while conservative individuals, in addition to concern for individuals, emphasize group-level morality: ingroup loyalty, respect for authority, and avoidance of impurity (e.g., sanctity, chastity).
These associations between moral values and politics, in particular between conservatism and increased concern about sexual purity, might suggest that conservative political orientation is associated with harsher judgment of people accused of sexual misconduct (Graham et al., 2009). However, the relevant existing research indicates that conservative individuals may have a tendency to judge sexual misconduct more leniently (e.g., Anderson et al., 1997; Lambert & Raichle, 2000), and attribute more responsibility and blame to victims (Niemi et al., 2020; Niemi & Young, 2016). The political orientation of perceivers, therefore, is plausibly hypothesized to influence moral evaluations of sexual misconduct.
The Influence of the Accused’s Political Orientation
The existence of ingroup bias has been extensively documented (e.g., Cikara, Botvinick, & Fiske, 2011; Cikara, Bruneau, & Saxe, 2011; Ellemers et al., 2019; Leach et al., 2007). There is accumulating evidence regarding the phenomena’s neurological underpinnings (Amodio & Cikara, 2021; Cikara, Botvinick, & Fiske, 2011; Molenberghs, 2013; Van Bavel et al., 2008), and the effect has been found for a variety of group characteristics, such as gender (Rudman & Goodwin, 2004), status (Bettencourt et al., 2001), wealth (Burkholder et al., 2021), and race (Depew et al., 2017; Whitehead et al., 1982). While ingroup bias is typically associated with socially salient identities, especially striking is that it can be experimentally induced even with manifestly arbitrary “minimal groups” (Brewer, 1979; Jackson et al., 2019). Like other phenomena, ingroup bias is likely to be highly contextual (Scheepers et al., 2006), but there is little question that it reliably appears in the domain of politics (Ehrlich & Gramzow, 2015; Finkel et al., 2020; Nicholson et al., 2016). However, while there is evidence that Republicans, compared to Democrats, are less likely to view sexual misconduct as a societal problem (Panagopoulos & van der Linden, 2018), there is relatively little documentation of ingroup effects in people’s responses to cases of sexual misconduct.
Research on the “Black Sheep Effect,” where ingroup transgressors are condemned more than outgroup transgressors, might be thought to imply that people may judge those accused of sexual misconduct more harshly when they share the accused person’ politics (Marques et al., 1988). Alternatively, people may show lenience toward perpetrators if it protects the group and existing social structures. In that case, liberals and conservatives alike may defend perpetrators and impugn victims, if it protects the group with which they align.
There is one study providing some systematic support for the possibility of a politicized ingroup effect on judgment of sexual misconduct. Klar and McCoy (2021), using a fictional scenario depicting a politician accused of sexual harassment, found that both Democrats and Republicans were less likely to judge an accused individual as “guilty” when the accused was a member of their own political party. As we shall see, the results of the present research are broadly in line with that finding; however, they go beyond evidence of political ingroup bias and use substantially different methods: to increase ecological validity, we used actual news stories in addition to fictional scenarios, and our dependent variables were in keeping with what have long been central concerns in moral psychology, including causation, responsibility, blame, and punishment (e.g., Alicke, 2000; Alicke et al., 2015; Cushman, 2008; Greene & Haidt, 2002; Shaver, 1985; Woolfolk et al., 2006; Young & Saxe, 2008). As such, the present research makes novel contributions to several literatures including research on sexual misconduct, ingroup bias, as well as moral and political psychology.
Hypotheses
In this research, we investigate how the moral judgment of sexual misconduct is influenced by three factors, the political orientation of media outlets, the personal political orientation of the participants assessing the cases, and the political orientation of the accused individual. Participants read news media reports (Studies 1-2) and fictional stimulus materials (Studies 3-4) about sexual misconduct, and rated the alleged perpetrators and victims on several variables including blameworthiness, responsibility, causal contribution (Studies 1-4), liability for punishment, and exploratory workplace-relevant variables (Studies 3-4).
These studies are framed by three overarching hypotheses:
Hypothesis 1: (Studies 1-2) Political orientation of news media outlet influences judgments of alleged perpetrators and victims, such that articles from conservative, relative to liberal, news outlets increase participants’ ratings of alleged victims as more likely, and alleged perpetrators as less likely, to be responsible, blameworthy, and causally-implicated in the alleged sexual misconduct.
Hypothesis 2: (Studies 1-4) Political orientation of participants influences judgments of alleged perpetrators and victims, such that conservative, relative to liberal participants, attribute more responsibility, blame, causation, and liability to punishment to alleged victims, and less to alleged perpetrators.
Hypothesis 3: (Studies 2 & 4) The relationship between participants’ political orientation and the political orientation of alleged perpetrators influences judgments of alleged perpetrators and victims. When participants share the same political orientation as the accused (relative to believing the political orientation of the accused differs from their own), participants’ judgments of responsibility, blame, causation, and punishment will favor the alleged perpetrator and disfavor the alleged victim.
Transparency and Openness
This research was conducted using open research methods. We report all measures, manipulations, and exclusions in all studies. Exclusions were based on the same attention check across the studies (see “Measures” in the Study 1 Method). For all studies, sample size was determined before data analysis, and data were analyzed only after the required sample size target was met. The data for the studies are linked from the Online Appendix. All studies reported in this paper were approved by the lead author’s university Institutional Review Board in the United States.
Study 1
Method
Participants
A total of 1,755 participants from the United States completed the study on Amazon’s Mechanical Turk (Mage = 38 years, SD = 12, age range 18 to 81; 45.0% men, 54.6% women, 0.4% chose “other”) for a small payment ($0.40). The eligibility criteria consisted of being over the age of 18, residing in the United States, and passing the attention check embedded at the end of the survey (see “Measures” below); 52 additional participants were excluded for failing this attention check (3.0% of participants who submitted the HIT). An additional 147 participants initially accepted the HIT but did not submit it (7.5% of all participants who accepted the HIT). Among those participants who completed the study and passed the attention check at the end, 62.6% reported having a Bachelor’s degree or higher. Participants’ mean political orientation (1 = very liberal to 7 = very conservative) was M (SD) = 3.69 (.05).
Procedure
The study consisted of a single, self-paced session administered via Amazon Mechanical Turk; on average it took participants less than 7 minutes to complete. Participants were randomly assigned to the liberal or conservative news source condition, then to one of three news sources in that condition, then to one of 20 articles from that news source (see “Stimulus Materials”). After reading the article, participants were presented with three sets of questions measuring their judgments of the accused person and accuser; a demographics survey followed (see “Measures”). Upon completion, participants received monetary compensation through Amazon’s Mechanical Turk for their time.
Stimulus Materials
We selected articles covering 20 different sexual misconduct allegations in the #MeToo period, beginning with the widely covered case of Harvey Weinstein (all 20 cases are listed in the Online Appendix Table S4). We included cases involving people from a range of professions, including politics, business, sports, and entertainment. In accordance with extensive survey results from the Pew Research Center (2014), we operationalized “liberal news outlets” as outlets distrusted by conservative people but trusted by liberal people, and “conservative news outlets” as outlets distrusted by liberal people but trusted by conservative people (Mitchell et al., 2014). Thus, the categorization of a news outlet as liberal or conservative in our study is directly informed by the perceptions of people self-identifying as liberal and conservative, respectively.
Constraining our choice of news outlets was a requirement that each of the outlets covered each of the 20 different cases we selected, so that we could examine data for each outlet for each of the 20 cases. We also restricted our choices to six total news outlets – three news outlets per political leaning, to control the cost of the study. The New York Times, Washington Post, and Huffington Post are all trusted by liberal but not conservative people, rendering these three news organizations “liberal news outlets”; in contrast, Breitbart, Fox News, and The Blaze are all trusted by conservative but not liberal people, rendering these “conservative news outlets.” Therefore, our article set consisted of 6 articles for each of 20 cases (120 total articles). Articles were not edited for content or length, but source information was redacted (see Online Appendix “Methodological Details: News outlets”).
Measures
After reading the article, participants were asked to read three sets of questions in a fixed order that measured (1) responsibility (“How much do you think each individual is responsible for the incident?”), and (2) causal contribution (“To what extent could a change in each individual's actions have changed the outcome?”) for the “Accuser(s)” and “[Last Name of Accused]” in counterbalanced order, using a Likert-scale (1 = not at all to 7 = very much), as well as (3) blame (“How much blame does [Last Name of Accused] deserve and how much blame do the accuser(s) deserve? Please designate a percent of blame for each individual to total 100.”)
Finally, participants provided demographic information (age, gender, education, ethnicity, race) and political orientation. We measured political orientation with the item (Iyer et al., 2012): “When it comes to politics, do you usually think of yourself as liberal, moderate, conservative, or something else?” – a drop-down menu contained the choices: (1) Very liberal, (2) Liberal, (3) Slightly liberal, (4) Moderate / middle-of-the-road, (5) Slightly conservative, (6) Conservative, (7) Very conservative, (8) ‘‘Don’t know / not political’’, (9) ‘‘Libertarian,’’ (10) Other. We used selections 1 – 7 (94.8% of selections) as a scale variable representing the extent of participants’ self-identification as politically liberal or conservative. Participants were also asked follow-up questions about prior knowledge; fewer than 1% of participants correctly identified the news outlet that published the article they read. We then presented participants with the following attention check: “Do you feel that you paid attention, avoided distractions, and took the survey seriously?” Participants selected one of the following: (1) no, I was distracted; (2) no, I had trouble paying attention; (3) no, I didn’t take the study seriously; (4) no, something else affected my participation negatively; or (5) yes. Participants were assured that their responses would not affect their payment or their eligibility for future studies. Only those participants who selected (5) were included in the analyses. The same attention check question has been used in previous research (e.g., Stanley, Marsh, & Kay, 2020; Stanley, Whitehead, et al., 2020; Stanley et al., 2019).
Data Analysis Plan
Our central statistical analyses involve linear mixed-effects models with crossed random-effects (see Supplementary Materials); at the time of writing, no accepted means of conducting a priori power analyses for such models was available. We therefore used a sample size based on the number of stimuli items (120 articles, aiming to obtain approximately 15 participants reading and rating each unique news article, after exclusions). Our data analysis plan involved using R (R Core Team, 2021) with the ‘lme4’ software package (Bates et al., 2014) to fit the data to linear mixed-effects models (LMEMs). To test significance of fixed-effects we used Satterthwaite approximations to degrees of freedom; 95% confidence intervals around beta-values were computed using parametric bootstrapping (number of simulations = 1000). On our view, 95% CIs around beta-values offer the best available indication of effect size for mixed-effects models with crossed random-effects. See Baayen et al. (2008) for a discussion of the benefits of using a mixed-effects modeling approach instead of more traditional approaches.
Results
To address our first hypothesis (H1) that the political leaning of the news outlets would predict judgments of responsibility, blame, and causation for accusers (alleged victims) and the accused (alleged perpetrators) featured the news articles, we computed six separate LMEMs in which the accused and the news outlet were included as crossed random-effects in all models (random intercepts only). The political orientation of the news outlet was modeled as a binary fixed-factor in all models. The political orientation of the news outlet did not significantly predict participants’ judgments of responsibility, blame, or causal contribution for the accused or the accuser. See “Study 1: News outlet” in the Online Appendix for these non-significant results, plus additional analyses.
Next, we addressed our second hypothesis (H2): the political orientation of the participant predicts judgments of responsibility, blame, and causation for the accusers and the accused. In six separate LMEMs, the accused and the news outlet were included as crossed random-effects (random intercepts only), and the reported political orientation of the participants (1 = very liberal to 7 = very conservative) was modeled as a fixed-effects term.
Results (see Table 1) showed that the political orientation of the participants significantly predicted judgments of the responsibility, blame, and causation for the accuser and the accused. Conservative participants, relative to liberal participants, judged the accuser to be significantly more responsible and the accused to be less responsible for the incident. Conservative participants, relative to liberal participants, assigned more blame to the accuser – which meant they assigned less blame to the accused by default. Finally, conservative participants, relative to liberal participants, rated the causal contribution of the accuser higher and the accused person lower.
Table 1
The Results of Six Different Linear Mixed-Effects Models in Studies 1-2 With the Political Orientations of Participants Predicting Judgments of Responsibility, Blame and Causation
| Study | Mlib | Mcons | b | SE | t | p | 95% CI |
|---|---|---|---|---|---|---|---|
| Responsibility | |||||||
| Accuser | |||||||
| 1 | 2.21 | 3.67 | -1.50 | 0.10 | -15.21 | < .0001 | [-1.70, -1.30] |
| 2 | 2.18 | 3.06 | 0.24 | 0.02 | 10.59 | < .0001 | [0.20, 0.29] |
| Accused | |||||||
| 1 | 6.24 | 5.76 | 0.51 | 0.07 | 7.47 | < .0001 | [0.38, 0.65] |
| 2 | 6.12 | 5.64 | -0.13 | 0.02 | -7.01 | < .0001 | [-0.17, -0.09] |
| Blame | |||||||
| Accuser | |||||||
| 1 | 13.49 | 30.88 | -17.85 | 1.30 | -13.75 | < .0001 | [-20.29, -15.24] |
| 2 | 14.07 | 25.11 | 3.08 | 0.34 | 9.04 | < .0001 | [2.41, 3.76] |
| Accused | |||||||
| 1 | 86.51 | 69.12 | 17.85 | 1.30 | 13.75 | < .0001 | [15.24, 20.29] |
| 2 | 85.93 | 74.89 | -3.08 | 0.34 | -9.04 | < .0001 | [-3.76, -2.41] |
| Causation | |||||||
| Accuser | |||||||
| 1 | 3.04 | 4.44 | -1.40 | 0.11 | -12.92 | < .0001 | [-1.62, -1.18] |
| 2 | 3.22 | 4.01 | 0.25 | 0.03 | 8.73 | < .0001 | [0.19, 0.30] |
| Accused | |||||||
| 1 | 6.14 | 5.80 | 0.34 | 0.08 | 4.23 | < .0001 | [0.18, 0.48] |
| 2 | 6.19 | 5.74 | -0.11 | 0.02 | -5.48 | < .0001 | [-0.15, -0.07] |
Note. Means are provided for liberal and conservative participants, respectively, for each outcome measure after collapsing across the specific news outlets and accused persons. All 95% CIs are for the beta-estimates. Accuser and accused blame ratings in Studies 1-2 are not independent because they were collected simultaneously by having participants assign a portion of a total of 100% of blame to both.
Study 2
In Study 2, we attempted to replicate our findings from Study 1 and then addressed our third hypothesis, that people who share the political orientation of the accused (relative to those who do not), will favor the accused, at the expense of the accuser(s), in judgments of responsibility, blame, and causation.
Method
Participants
A new sample of participants (N = 1,718) from the United States completed the study on Amazon’s Mechanical Turk (Mage = 37 years, SD = 11, age range 18-78) for a small payment ($0.40). Participants were 55.8% men, 43.7% women, .004% chose “other”; 63.9% of participants reported having a Bachelor’s degree or higher. Participants indicated their own political orientation by selecting (1) liberal, (2) conservative, (3) other, or (4) I don’t know, rather than using a Likert scale as in Study 1 (this change was intended to make it easier to interpret and decompose possible interaction effects within the mixed-effects modeling framework). A total of 839 participants reported being liberal, 601 participants reported being conservative, 249 participants reported other, and 29 reported I don’t know. Eligibility criteria were the same as in Study 1; 96 additional participants were excluded for failing the attention check (5.6% of participants who submitted the HIT; see “Measures,” Study 1). An additional 136 participants initially accepted the HIT but did not submit it (7.0% of all participants who accepted the HIT).
Procedure and Stimulus Materials
The procedure and stimulus materials in Study 2 were the same as in Study 1, with the addition of a measure of perceived political orientation of the accused. The study took participants less than 7 minutes on average.
Measures
The measures in Study 2 were the same as in Study 1, with two exceptions: first, the measure of participant politics, as described in the “Participants” section above, and, second, the perceived political orientation of the accused individuals, for which participants selected from the following options: (1) liberal, (2) conservative, (3) other, or (4) I don’t know. Participants considered 656 accused individuals conservative and 543 liberal. They selected “I don’t know” 480 times and “other” 39. Of the participants who perceived the politics of the accused person to be liberal or conservative, being liberal or conservative did not appear to affect their perceptions of the accused person’s politics (chi-square test p = .095). The data of participants who were liberal or conservative who rated an accused individual liberal or conservative was used for these analyses (n = 485 conservatives, n = 579 liberals, total N = 1,064).
Data Analysis Plan
The same data analysis plan used in Study 1, including the statistical methods and software, was used in Study 2.
Results
First, replicating Study 1, political orientation of the news outlet did not significantly predict responsibility, blame, or causation judgments, therefore (H1) was again not supported in Study 2. See “Study 2: News outlet” in the Online Appendix for the full reporting of these non-significant results.
Next, also replicating Study 1, the political orientations of the people assessing the cases predicted judgments of responsibility, blame, and causation for the accusers and the accused (see Figure S2 in the Online Appendix). Conservative, relative to liberal, participants judged the accuser to be more responsible, blameworthy, and causal, and the accused to be less responsible, blameworthy, and causal for the incident (Table 1).
Finally, we addressed our third hypothesis (H3): the perceived political orientation of the accused person influences the effect of the participant’s political orientation on their judgments of responsibility, blame, and causation. To this end, we computed six separate LMEMs in which the reported political orientation of the participants and the perceived political orientation of the accused (both: liberal coded as 1, conservative coded as 0) were modeled as fixed-effects in all models; and the accused and the particular news outlet were included as crossed random-effects, in all models (random intercepts only). Collectively, the results indicate a consistent, significant interaction between the political orientation of the participant and the political orientation of the accused in predicting judgments of responsibility, blame, and causation for both the accusers and accused (see Table 2).
Table 2
Results of Six Different Linear Mixed-Effects Models From Study 2, With Participant and Accused Political Orientation as Fixed Predictors of Each of the Outcome Measures
| Outcome measures | b | SE | t | p | 95% CI |
|---|---|---|---|---|---|
| Responsibility | |||||
| Accuser | |||||
| Intercept | 4.61 | 0.15 | 30.46 | < .0001 | [4.30, 4.91] |
| Participant Politics | -2.43 | 0.15 | -16.10 | < .0001 | [-2.71, -2.13] |
| Accused Politics | -1.73 | 0.17 | -10.20 | < .0001 | [-2.07, -1.38] |
| Participant Politics x Accused Politics | 2.09 | 0.23 | 9.23 | < .0001 | [1.60, 2.52] |
| Accused | |||||
| Intercept | 5.49 | 0.10 | 53.86 | < .0001 | [5.30, 5.69] |
| Participant Politics | 0.82 | 0.11 | 7.73 | < .0001 | [0.61, 1.02] |
| Accused Politics | 0.61 | 0.12 | 5.09 | < .0001 | [0.36, 0.84] |
| Participant Politics x Accused Politics | -0.73 | 0.16 | -4.06 | < .0001 | [-1.05, -0.40] |
| Blame | |||||
| Accuser | |||||
| Intercept | 41.14 | 1.82 | 22.64 | < .0001 | [37.37, 44.59] |
| Participant Politics | -28.42 | 1.97 | -14.42 | < .0001 | [-32.25, -24.27] |
| Accused Politics | -20.53 | 2.23 | -9.21 | < .0001 | [-25.12, -16.06] |
| Participant Politics x Accused Politics | 24.50 | 2.95 | 8.30 | < .0001 | [18.02, 30.16] |
| Accused | |||||
| Intercept | 58.87 | 1.82 | 32.40 | < .0001 | [55.28, 62.24] |
| Participant Politics | 28.42 | 1.97 | 14.42 | < .0001 | [24.73, 32.46] |
| Accused Politics | 20.53 | 2.23 | 9.21 | < .0001 | [16.25, 24.91] |
| Participant Politics x Accused Politics | -24.50 | 2.95 | -8.30 | < .0001 | [-30.53, -18.90] |
| Causation | |||||
| Accuser | |||||
| Intercept | 5.08 | 0.14 | 37.07 | < .0001 | [4.82, 5.36] |
| Participant Politics | -2.10 | 0.16 | -12.75 | < .0001 | [-2.42, -1.78] |
| Accused Politics | -1.20 | 0.19 | -6.47 | < .0001 | [-1.56, -0.83] |
| Participant Politics x Accused Politics | 1.55 | 0.25 | 6.27 | < .0001 | [1.09, 2.03] |
| Accused | |||||
| Intercept | 5.55 | 0.11 | 51.69 | < .0001 | [5.35, 5.76] |
| Participant Politics | 0.58 | 0.12 | 4.69 | < .0001 | [0.32, 0.82] |
| Accused Politics | 0.57 | 0.14 | 4.06 | < .0001 | [0.29, 0.85] |
| Participant Politics x Accused Politics | -0.64 | 0.19 | -3.41 | .0007 | [-0.99, -0.27] |
Note. All 95% CIs are for the beta-estimates. For both political orientation variables, conservative was coded as 0, and liberal was coded as 1.
To further examine the significant interaction effects in the LMEMs, follow-up linear mixed-effects models were computed for conservative and liberal participants, separately. The results indicated that conservative participants tended to judge liberal accused people, relative to conservative accused people, to be more responsible, b = .68, SE = .13, t = 5.46, p < .0001, 95% CI [.44, .92], more blameworthy, b = 20.79, SE = 2.32, t = 8.95, p < .0001, 95% CI [16.68, 25.22], and more causally-implicated, b = .62, SE = .13, t = 4.69, p < .0001, 95% CI [.37, .87]; and the accusers as less responsible, b = -1.79, SE = .18, t = -9.98, p < .001, 95% CI [-2.15, -1.41], less blameworthy, b = -20.79, SE = 2.32, t = -8.95, p < .001, 95% CI [-25.22, -16.08], and less causally-implicated, b = -1.20, SE = .18, t = -6.79, p < .001, 95% CI [-1.54, -.86]. Symmetrically, liberal participants tended to judge liberal accused people, relative to conservative accused people, to be less blameworthy, b = -4.02, SE = 1.96, t = -2.05, p = .041, 95% CI [-7.54, -.02], and their victims more responsible, b = .36, SE = .15, t = 2.47, p = .014, 95% CI [.09, .65], and more blameworthy, b = 4.02, SE = 1.96, t = 2.05, p = .041, 95% CI [.35, 8.06]. However, liberal participants did not judge the responsibility of the accused, b = -.03, SE = .11, t = -.30, p = .76, 95% CI [-.25, .19], or the causal contribution of the accused, b = .01, SE = .13, t = .07, p = .95, 95% CI [-.26, .27], or their accusers, b = .26, SE = .16, t = 1.61, p = .11, 95% CI [-.06, .57], differently based on whether they believed the accused to be liberal or conservative. To facilitate interpretation of these data, conservative and liberal participants’ mean ratings of responsibility, blame, and causation based on the perceived politics of the accused individuals (data collapsed across accused and news outlet) are provided in Table 3 and illustrated in Figure 1.
Table 3
Conservative and Liberal Participants’ Mean Ratings of Responsibility, Blame, and Causation Based on the Perceived Politics (Conservative or Liberal) of the Accused in Study 2
| Perceived politics of accused | Mconservative | Mliberal | F | p | |
|---|---|---|---|---|---|
| Responsibility | |||||
| Accuser | |||||
| conservative | 4.56 | 2.12 | 103.46 | < .001 | .09 |
| liberal | 2.84 | 2.56 | 8.36 | .004 | .01 |
| Accused | |||||
| conservative | 5.47 | 6.27 | 33.66 | < .001 | .03 |
| liberal | 6.16 | 6.21 | 0.30 | .59 | .00 |
| Blame | |||||
| Accuser | |||||
| conservative | 40.76 | 12.34 | 88.80 | < .001 | .08 |
| liberal | 20.00 | 16.90 | 5.22 | .02 | .01 |
| Accused | |||||
| conservative | 59.24 | 87.66 | 88.80 | < .001 | .08 |
| liberal | 80.00 | 83.10 | 5.22 | .02 | .01 |
| Causation | |||||
| Accuser | |||||
| conservative | 5.06 | 2.94 | 42.61 | < .001 | .04 |
| liberal | 3.87 | 3.33 | 5.37 | .02 | .01 |
| Accused | |||||
| conservative | 5.54 | 6.11 | 19.81 | < .001 | .02 |
| liberal | 6.16 | 6.10 | 0.01 | .91 | .00 |
Note. Mean ratings by conservative and liberal participants are provided for the accusers (alleged victims) and accused persons in two conditions that varied perceived politics of the accused person (conservative or liberal).
Figure 1
Conservative and Liberal Participants’ Judgments of Alleged Victims and Accused Persons in Studies 1 and 2
Note. Error bars indicate SEM. Alleged victims' and accused persons' blame ratings are not independent as participants assigned a total of 100% of blame to both. Data collapsed across accused persons and news outlets. Accused persons' politics were Unspecified in Study 1, and Conservative or Liberal in Study 2.
Study 3
In Study 3, we aimed to conceptually replicate our findings in Study 1, this time using fictional scenarios instead of real news articles. The use of actual news articles in Studies 1-2 underscored the importance of research on psychological processing of sexual misconduct using naturalistic, ecologically valid stimulus materials. Here, we aimed to isolate our variables of interest and test our hypotheses with controlled stimuli: simple fictional scenarios depicting sexual misconduct. The combination of these methodologies supports confidence that the findings, which are largely consistent across both kinds of stimulus materials, are both unconfounded and ecologically valid.
Method
Participants
A new sample of participants (total N = 354) from the United States completed the study on Prolific (Mage = 36, SD = 13, 48% men, 49% women, 3% chose “other”) for a small payment; 3 additional participants were excluded for failing an attention check embedded at the end of the survey. Of the total participants, 182 were political conservatives, and 172 were political liberals; because we recruited only conservatives and liberals, no participants were dropped from the analyses.
Materials and Procedure
After providing informed consent, all participants were presented with the following fictional scenario:
Steve is an executive vice president at a large corporation. Steve and his new administrative assistant, Jane, go on a work trip to one of the company’s production facilities. After they get to their hotel, Steve calls Jane to his room to give her some important company documents that need to be delivered to the production facility. When Jane arrives, she briefly chats with Steve about the day’s events. A few co-workers passing in the hallway hear them laughing. As Jane starts to leave, Steve reaches in and forcefully kisses her. Jane yells “Stop it!” and rushes from the room. After the trip is over, Jane files a formal company sexual harassment complaint against Steve.
After reading the scenario, participants made a total of 13 judgments. Similarly to the previous studies, participants completed six items indicating how responsible and blameworthy Jane and Steve were, respectively, for the incident (1 = not at all, 7 = completely); as well as how much Jane and Steve could have, respectively, changed the outcome of the incident (1 = not at all, 7 = completely). Participants also responded to four new items assessing punishments: participants indicated whether Jane or Steve should, respectively, be reprimanded by a superior for the incident (1 = definitely no, 7 = definitely yes), and whether Jane or Steve should, respectively, be fired for the incident (1 = definitely no, 7 = definitely yes).
Results
We first conducted a MANOVA with political orientation of the participant (liberal vs. conservative) and the six dependent measures from Studies 1-2, which revealed a significant main effect of political orientation, F(6,344) = 4.64, p < .001; Wilk's Λ = 0.93, = .08. Univariate follow up tests are reported in Table 4. As in Studies 1-2, political orientation was related to moral judgments such that conservative participants judged Jane, the alleged victim, as more responsible, blameworthy, and causal than liberal participants. Conservative participants judged Steve, the alleged perpetrator, no differently than liberal participants in terms of responsibility, blame, and causation judgments.
Table 4
The Results of Analyses in Studies 3-4 With the Political Orientations of Participants Predicting Judgments
| Target of judgment | Study | Mliberal | Mconservative | F | p | |
|---|---|---|---|---|---|---|
| Responsibility | ||||||
| Jane | 3 | 1.36 | 1.74 | 8.53 | .004 | .02 |
| 4 | 1.26 | 1.58 | 25.46 | < .001 | .03 | |
| Steve | 3 | 6.82 | 6.72 | 1.78 | .183 | .01 |
| 4 | 6.90 | 6.79 | 14.26 | < .001 | .02 | |
| Blame | ||||||
| Jane | 3 | 1.26 | 1.69 | 17.72 | < .001 | .05 |
| 4 | 1.28 | 1.55 | 16.90 | < .001 | .02 | |
| Steve | 3 | 6.82 | 6.75 | 1.16 | .282 | .00 |
| 4 | 6.86 | 6.78 | 5.23 | .022 | .01 | |
| Causation | ||||||
| Jane | 3 | 2.13 | 3.12 | 17.94 | < .001 | .05 |
| 4 | 2.09 | 2.77 | 44.69 | < .001 | .05 | |
| Steve | 3 | 6.79 | 6.71 | 1.28 | .260 | .00 |
| 4 | 6.76 | 6.66 | 3.26 | .071 | .00 | |
| Cooperate | ||||||
| Jane | 3 | 5.86 | 5.64 | 2.19 | .140 | .01 |
| 4 | 5.99 | 5.67 | 12.76 | < .001 | .01 | |
| Steve | 3 | 1.38 | 1.83 | 13.86 | < .001 | .04 |
| 4 | 1.41 | 1.81 | 26.79 | < .001 | .03 | |
| Reprimand | ||||||
| Jane | 3 | 1.27 | 1.66 | 12.02 | < .001 | .03 |
| 4 | 1.26 | 1.56 | 17.91 | < .001 | .02 | |
| Steve | 3 | 6.78 | 6.54 | 6.18 | .013 | .02 |
| 4 | 6.78 | 6.61 | 8.56 | .004 | .01 | |
| Fire | ||||||
| Jane | 3 | 1.13 | 1.28 | 3.60 | .059 | .01 |
| 4 | 1.13 | 1.22 | 3.39 | .066 | .00 | |
| Steve | 3 | 6.34 | 5.76 | 15.24 | < .001 | .04 |
| 4 | 6.33 | 5.86 | 30.92 | < .001 | .03 | |
In Study 3, we also included variables asking participants if they would cooperate with, reprimand, or fire Jane and Steve. A MANOVA with political orientation of the participant (liberal vs. conservative) and the six dependent measures revealed a significant main effect of political orientation, F(6,344) = 4.06, p < .001; Wilk's Λ = 0.93, = .07. Univariate follow up tests (Table 4) showed that conservative participants, compared to liberal participants, were more likely to rate Steve as someone with whom they’d cooperate, and as someone who did not deserve to be reprimanded or fired. Conservative participants, relative to liberal participants, rated Jane as more deserving of reprimand. In sum, conservative participants did not morally judge Steve differently from liberal participants, yet, they were more likely to endorse workplace leniency for Steve than liberal participants.
Study 4
In Study 4, we aimed to conceptually replicate our findings in Study 2, which demonstrated that conservative participants were more lenient in their moral judgments of alleged conservative perpetrators, at the expense of the victims. This time we used fictional scenarios instead of real news articles to test replication of the finding in more controlled conditions.
Method
Participants
A new sample of participants (total N = 1,001) from the United States completed the study on Prolific (Mage = 34 years, SD = 12, 43% men, 54% women, 3% chose “other”) for a small payment; 107 participants were excluded for failing a reading comprehension check or the attention check embedded at the end of the survey. Of the total, 499 participants were conservatives, and 502 participants were liberals; because we recruited only conservatives and liberals, no participants were dropped from the analyses.
Materials and Procedure
After providing informed consent, all participants were randomly assigned to read one of two scenarios in a between-subjects fashion. The scenarios were the same as in Study 3, but with one critical exception: a clause was added to the first sentence of each scenario to provide information implying Steve’s political orientation was either liberal or conservative. Participants assigned to the liberal perpetrator condition read that Steve was liberal; by contrast, participants assigned to the conservative perpetrator condition read that Steve was conservative.
After participants read the scenario to which they were assigned, participants completed a manipulation check, which asked what Steve’s political leaning was (binary response: liberal or conservative). Next, participants responded to two comprehension check questions, indicating whether, based on the scenario they read, the follow statements were true or false: Steve forcefully kissed Jane, and, Jane filed a sexual harassment complaint against Steve. Failing any one of these three items resulted in exclusion from our analyses. Subsequently, using the same 7-pt Likert-type scales, participants then made judgments on the same dependent variables as in Study 3. Finally, participants answered the demographics questions and responded to the same attention check question as in the previous studies. Upon completion, participants were monetarily compensated.
Results
Replicating Studies 1-3, participants’ political orientations predicted their judgments of responsibility, blame, and causation. Again, a MANOVA revealed a significant effect of political orientation (liberal coded as 1, conservative coded as 0; F(6,887) = 8.38, p < .001; Wilk's Λ = 0.95, = .05); univariate tests showed that conservative, relative to liberal, participants judged the accuser to be more responsible, blameworthy, and causal, and the accused to be less responsible and blameworthy (but not causal this time). See Table 4 for the results and means. We additionally asked participants if they would cooperate with, reprimand, or fire Jane and Steve. A MANOVA with political orientation of the participant (liberal vs. conservative) and the 6 dependent measures revealed a significant main effect of political orientation, F(6,887) = 8.84, p < .001; Wilk's Λ = 0.94, = .06. Univariate follow up tests are reported in Table 4. Conservative participants were more likely to cooperate with Steve; less likely to reprimand or fire Steve; more likely to reprimand Jane; and less likely to cooperate with Jane, relative to liberal participants.
In Study 4, we again addressed our hypothesis that the political orientation of the accused person influences the relationship between the participants’ political orientation and their judgments of responsibility, blame, and causation. We modeled participant politics (liberal vs conservative) and politics condition (liberal vs conservative) as fixed effects in a MANOVA with the six dependent measures responsibility, blame, and causation ratings for Jane and Steve. There was a significant effect of the participants’ political orientation, F(6,885) = 8.39, p < .001; Wilk's Λ = 0.95, = .05, no effect of politics condition, F(6,885) = 1.87, p = .082; Wilk's Λ = 0.99, = .01, and a significant interaction of participant politics and politics condition, F(6,885) = 2.57, p = .018; Wilk's Λ = 0.98, = .02. Table 5 depicts the results of the follow up tests for each of the dependent measures. The results conceptually replicated Study 2: participants’ political orientation interacted with the accused’s political orientation to predict moral judgments, this time with more precisely controlled stimulus materials. As seen in Table 5, conservative participants were more lenient towards the accused and less lenient towards the accuser – when the accused was apparently also conservative. This time, liberal participants were not more lenient toward an apparently liberal accused person or less lenient toward their accuser (p’s > .05); liberal participants’ ratings were consistent across the conservative and liberal accused conditions.
Table 5
The Results of Analyses Based on the Participants’ Political Orientation (Liberal, Conservative) and Politics Condition in Study 4
| Target of judgment | Condition | Mliberal | Mconservative | Mdiff | F | p | |
|---|---|---|---|---|---|---|---|
| Responsibility | |||||||
| Jane | conservative | 1.24 | 1.73 | .316 | 11.88 | < .001 | .013 |
| liberal | 1.27 | 1.42 | -.026 | .09 | .768 | .000 | |
| Steve | conservative | 6.93 | 6.72 | -.148 | 12.31 | < .001 | .014 |
| liberal | 6.87 | 6.87 | .057 | 2.01 | .157 | .002 | |
| Blame | |||||||
| Jane | conservative | 1.25 | 1.63 | .167 | 3.00 | .084 | .003 |
| liberal | 1.30 | 1.46 | -.048 | .27 | .606 | .000 | |
| Steve | conservative | 6.87 | 6.69 | -.181 | 10.94 | <.001 | .012 |
| liberal | 6.85 | 6.87 | .020 | .14 | .709 | .000 | |
| Causation | |||||||
| Jane | conservative | 2.08 | 3.01 | .479 | 10.69 | .001 | .012 |
| liberal | 2.10 | 2.53 | -.023 | .03 | .870 | .000 | |
| Steve | conservative | 6.76 | 6.59 | -.145 | 3.11 | .078 | .003 |
| liberal | 6.77 | 6.74 | -.005 | .00 | .954 | .000 | |
| Cooperate | |||||||
| Jane | conservative | 5.98 | 5.61 | -.366 | 8.47 | .004 | .009 |
| liberal | 5.99 | 5.72 | -.271 | 4.54 | .033 | .005 | |
| Steve | conservative | 1.32 | 1.93 | .611 | 30.67 | <.001 | .033 |
| liberal | 1.50 | 1.70 | .199 | 3.17 | .075 | .004 | |
| Reprimand | |||||||
| Jane | conservative | 1.23 | 1.63 | -.199 | 16.23 | <.001 | .018 |
| liberal | 1.29 | 1.49 | .397 | 3.78 | .052 | .004 | |
| Steve | conservative | 6.81 | 6.59 | -.218 | 6.49 | .011 | .007 |
| liberal | 6.75 | 6.62 | -.137 | 2.51 | .114 | .003 | |
| Fired | |||||||
| Jane | conservative | 1.15 | 1.30 | .155 | 5.85 | .016 | .007 |
| liberal | 1.12 | 1.13 | .011 | .03 | .864 | .000 | |
| Steve | conservative | 6.36 | 5.74 | -.620 | 27.15 | < .001 | .030 |
| liberal | 6.29 | 5.98 | -.318 | 6.99 | .008 | .008 | |
Note. Means are provided for liberal and conservative participants, respectively, for each outcome measure for the liberal and conservative conditions. Accuser and accused blame ratings are not independent because they were collected simultaneously by having participants assign a portion of a total of 100% of blame to both.
Finally, we conducted a MANOVA with participant politics (liberal vs conservative) and politics condition (liberal vs conservative) on the six workplace measures: cooperation, reprimand, and firing for Jane and for Steve. It revealed a significant main effect of political orientation, F(6,885) = 8.80, p < .001; Wilk's Λ = 0.94, = .06, and no main effect of politics condition or interaction. See means in Table 5. When the alleged perpetrator was portrayed as conservative, all 6 of the 6 dependent measures significantly differed across liberal and conservative participants. Conservative participants, relative to liberal participants, were more likely to cooperate with and less likely to reprimand or recommend firing Steve; they were less likely to cooperate with the alleged victim, and were more likely to reprimand and recommend firing her. When the alleged perpetrator was liberal, just 2 of the 6 dependent measures significantly differed across liberal and conservative participants: ratings of cooperation with Jane, and interest in firing Steve.
General Discussion
When then Supreme Court nominee Brett Kavanaugh was accused of sexual assault in 2018, 52% of Republicans thought the sexual assault, if substantiated, would not disqualify Kavanaugh from serving on the highest Court, compared to 14% of Democrats (The Economist/YouGov Poll, 2018). Previous research suggested that polarized reactions to sexual misconduct allegations are at least partly due to individual differences, yet there has been little empirical, scientific attention given to news consumers’ politics and their perceptions of alleged offenders’ politics on their moral judgments about cases of sexual misconduct.
We presented here four studies investigating the influence of political orientation on moral judgments about cases of sexual misconduct. First, our hypothesis that political orientation would influence attributions of responsibility, blame, causation, and punishment for alleged perpetrators and victims was supported. Conservative, relative to liberal participants, attributed more responsibility, blame, causation, and punishment to alleged victims, and less to alleged perpetrators, across the four studies (H2). In these studies, which used both actual news stories (Studies 1-2) and fictional scenarios describing sexual misconduct (Studies 3-4), conservative participants were more likely to consider the people making the accusations of sexual misconduct to be responsible for the incident, to believe that these accusers could have changed the outcome, to blame accusers for what transpired, and to consider the accuser deserving of reprimand. Moreover, conservative participants were less likely to hold the accused individual responsible, to believe that the accused could have changed the outcome, and to blame the accused for what transpired. They were also less likely to support reprimanding or firing the accused individual.
These results indicate that conservative participants generally demonstrated an increased tendency for “victim-blaming” and leniency toward alleged perpetrators. This accords with previous research showing that conservative and liberal individuals in the United States differ in their attribution of blame and responsibility in response to sexual misconduct: conservatives and those higher in moral values associated with conservativism more often perceive victims as deserving of blame for the assault and as causing the assault (e.g., Anderson et al., 1997; Lambert & Raichle, 2000; Niemi & Young, 2016; Niemi et al., 2020).
Our hypothesis that the political orientation of participants, together with the political orientation of alleged perpetrators, influences judgments of alleged perpetrators and victims (H3) was also supported. We hypothesized that when participants shared the same political orientation as the accused (relative to believing the political orientation of the accused differed from their own), their moral judgments would favor the alleged perpetrator and disfavor the alleged victim. Indeed, our results showed that participants gave more exculpating judgments of responsibility, blame, causation, and punishment to alleged perpetrators (and more inculpating judgments to alleged victims) when they shared the political orientation of the alleged perpetrator.
What about the Americans who prefer the “middle of the road” approach? Are our results concerning liberals and conservatives somehow affected by excluding them? In the first study, we do not exclude them; political orientation was measured as a scale variable ranging from 1-7. To explore whether dichotomizing politics by excluding the “middle-of-the-road/moderate” participants substantially affects the results, we ran “Additional Analysis of Participant Politics” on the Study 1 data (see Online Appendix) and follow up tests with liberal (n = 800, coded 1 = 1, 2, 3 on the scale) and conservative (n = 566, coded 0 = 5, 6, 7) entered to predict each of the six dependent measures. The effects of politics on responsibility, blame, and causation for the accuser and the accused replicate those reported for Study 1 with the scalar variable, with essentially identical p’s < .001 (see Table S3 in the Online Appendix). This strongly suggests the loss of the midpoint participants did not affect the main results. Furthermore, the results of Study 2, where we collected political orientation using a categorical variable (liberal, conservative, other, or “I don’t know” options), closely replicates Study 1 where politics was measured with a scale variable.
We acknowledge that simple dichotomized measures of political orientation have limitations, yet when the purpose is to study liberals and conservatives, as ours was, they are valid measures. The dichotomy also aptly fits with the present picture of life in the US: there is accumulating evidence that Americans, like those in many other nations, have become increasingly polarized: more likely to dislike, distrust, and even vilify their political opponents, attributing negative traits and qualities to them (e.g., unintelligent, hypocritical, selfish; Iyengar et al., 2012; Abramowitz & Webster, 2016; Levendusky & Malhotra, 2016; Klar & McCoy, 2021; Reiljan, 2020; Wagner, 2021). Evidently, moral judgments of both victims and perpetrators are likely to depend on whether the individual being judged is thought to be on the same political “team,” rather than on consistently and impartially applied moral principles.
However, this politicized ingroup bias in responses to sexual misconduct was reliable and more pronounced in conservative participants, compared to liberal participants. It is possible that when judging cases of sexual misconduct, conservatives’ moral concerns related to sexual misconduct were overridden by other important conservative values (Graham et al., 2009), in particular, a “binding” loyalty to the political ingroup. Alternatively, a concern for “personal responsibility” that characterizes conservative ideology may have produced the victim-blaming observed. This seems unlikely since when conservatives judged a fellow conservative accused person, they not only attributed more responsibility to the victim, they were also more likely to withhold it from the accused individual. This leniency toward politically-aligned alleged perpetrators suggests that a general conservative tendency to assign more personal responsibility is unlikely to explain the results, whereas a tendency to target moral judgments in order to favor the ingroup is a more likely explanation. Varying the political orientation of the victim in future work would help clarify whether being in the victim role or being politically aligned was more consequential to conservatives’ and liberals’ moral judgments.
By contrast, traditionally liberal moral concerns about impartiality and “universal” moral concern directly oppose conservative “binding” values (Graham et al., 2011). It is possible that when liberal participants judged sexual misconduct, their valuation of impartiality reduced their tendency to demonstrate loyalty or favoritism to alleged liberal perpetrators. Relatedly, it is possible that a preference for impartiality and a desire to blame perpetrators of sexual misconduct, resulted in liberal participants’ judgments artificially exhibiting little variability and remaining at the high end of the scale. Regardless of the exact mechanism, it was the case that more conservative participants were consistently more likely to make judgments characterized by political favoritism that shielded politically-aligned accused parties from blame (Studies 2 & 4), and shifted blame to the accusers (All studies). Ingroup favoritism alone cannot explain these results, as it would predict that liberals’ and conservatives’ judgments would be politically skewed to a similar extent.
Future research should explore whether partisan moral judgments are also found for non-sexual offenses like financial fraud, mugging, or identity theft. If participants judge politically-aligned offenders more leniently, and their victims more harshly, for non-sexual and sexual offenses alike, this would suggest a broader tendency to protect ingroup offenders and delegitimize their victims, across transgression type. Some prior work supports this possibility in the case of conservative politics: participants higher in conservative moral values, compared to lower, judged victims of both rape and robbery as more culpable (Niemi & Young, 2016).
Alternatively, if the effect on conservatives’ judgments is specific to sexual transgressions (and still minimal for liberals), this would suggest that there is something unique to sexual transgressions and conservative ideology that motivates enhanced leniency for alleged perpetrators and harsher judgments of alleged victims. Some research supports this possibility: conservative values like purity and ingroup loyalty, belief in rape myths, sexual aggression, and hostile sexism are difficult to empirically extricate from each other — this clustering of correlated attitudes and beliefs may explain conservatives’ more favorable judgments of conservative perpetrators of sexual misconduct (Begany & Milburn, 2002; Koss & Dinero, 1988).
Finally, these results build on broader work demonstrating that people’s evaluations of new stimuli are strongly dependent upon their prior beliefs and biases (Kunda, 1990; Lodge & Taber, 2013; Stanley, Henne, et al., 2020). In the moral domain, people rarely change their minds after exposure to evidence against their priors (Stanley et al., 2018; Uhlmann et al., 2009). The present research extends these lines of research by shedding new light on the role of people’s political beliefs in making moral judgments about real-world events. The results reveal that people bring their established political beliefs and biases to their moral evaluations of new information, which can exert considerable influence on their judgments, sometimes crowding out moral principles to which they explicitly attest.
Interestingly, our hypothesis that the political orientation of news media outlets influences judgments of responsibility, blame, and causation for alleged perpetrators and victims was not supported (H1). The political orientation of news outlets did not affect any of the measured judgments for either alleged perpetrators or victims. This result is surprising, given partisan claims of “fake news” or misinformation – whether people can expect fair or accurate reporting from major news outlets is frequently questioned, and has even led to news outlets being excluded from covering events of concern to the public at the national level.
Limitations and Future Directions
Past work suggests that there are gender differences in judgments of the victims and perpetrators of sexual assault, such that men tend to engage in more victim-blaming than women (Grubb & Turner, 2012; Hayes et al., 2013). In this research, the role of participant gender, age, and other aspects related to identity (e.g., surviving sexual assault) were not explored. While there is no reason to think these factors systematically vary with political orientation in ways that would confound our research design, they may intersect with how individuals consume accounts of high-profile cases of sexual misconduct. Other demographic factors we did not explore might also affect how news stories about sexual misconduct are consumed, such as race, ethnicity, and socioeconomic status.
Online samples represent convenience samples and are not representative of the general population of the United States (e.g., they are typically more liberal and less ethnically diverse). Nevertheless, online samples may outperform other sources of data, such as college student samples, with respect to diversity in socio-economic and political characteristics of interest (Berinsky et al., 2012; Buhrmester et al., 2011; Hauser & Schwarz, 2016; Mullinix et al., 2015). In addition, similar results have been observed in online and in-lab testing environments for experiments used to study established psychological phenomena with adults (e.g., Stroop effect, Coppock, 2019) and children (Chuey et al., 2022), indicating the quality of online data is broadly comparable to data collected with traditional methods.
Consistent with the dramatically polarized political climate in the United States (Abramowitz & Saunders, 2008; Pew Research Center, 2014), the current results indicate that evaluations of cases of sexual misconduct are likely to be influenced by the match between the politics of the evaluator and the politics of the accused individual. But is the influence of politics on moral judgments of sexual misconduct unique to the United States? Some work suggests the current findings would be replicated in other Western and non-Western countries. More and more, the global political landscape is looking similar to the polarized liberal-conservative dichotomy of the US (Reiljan, 2020; Wagner, 2021). Increasing left-right polarization outside the US suggests that the effects would manifest in non-US samples. Nevertheless, country- and region-level differences in attitudes about gender roles, domestic violence, and sexual autonomy are also present (World Health Organization, 2021). Future work should explore whether and how the effects of politics on moral judgment are influenced by cultural factors.
Political favoritism for alleged perpetrators of sexual misconduct was pronounced in people identifying as conservative, yet it was observed in both liberal and conservative participants. These findings open several avenues for future research; for example, analyses might reveal the extent to which politics predict moral evaluations of alleged perpetrators and victims alongside factors related to demographics, personal history, sexism, and rape myth acceptance. Research that clarifies the beliefs and values underlying identification with conservative and liberal political ideology, and the contexts in which they emerge, is needed to increase understanding of politicized judgment of sexual misconduct, and its morally-relevant consequences.