Public Opinion on Sexual Harassment in Politics

To develop our research expectations, we first consider the growing body of scholarship on political misconduct and SH. In these studies, partisanship and party loyalty have emerged as key predictors of voters’ willingness to punish candidates who engage in this behavior (Cossette and Craig Reference Cossette and Craig2020; Hansen and Dolan Reference Hansen2020; Craig and Cossette Reference Craig and Cossette2022; Stark and Collignon Reference Stark and Collignon2022). Cossette and Craig (Reference Cossette and Craig2020)’s U.S. study finds that Democrats are more likely than Republicans to report feeling sad and angry that SH occurs in politics and to say that victims who claim to have been sexually harassed should be taken at their word. Looking at partisanship and gender, Masuoka, Grose and Junn (Reference Masuoka, Grose and Junn2021) find that Democratic women are the most likely, and Republican men the least likely, to electorally punish a congressional candidate accused of SH. What remains unknown, however, are the factors that condition peoples’ attitudes about SH in politics in contexts where political parties do not operate, which is the case in many municipal contexts around the world.

Women’s and Men’s Attitudes About Sexual Harassment

In the absence of party leadership cues, we suspect that women will be especially attuned to this problem and see a compelling need to sanction councillors who engage in this behavior compared to men. Prior research demonstrates, for example, that women are motivated by campaign dynamics when gender-salient issues are at stake (Hansen and Dolan Reference Hansen2020; Cassese and Barnes Reference Cassese and Barnes2018). Although men experience SH, women are more likely to be sexually harassed, which we argue makes it a highly gender-salient issue (Uggen and Blackstone Reference Uggen and Blackstone2004). Compared to men, women are also more likely to describe a wider range of behaviors as SH and are generally less tolerant of such behavior (Fitzgerald and Shullman, Reference Fitzgerald and Shullman1993).

Research on sexual misconduct in politics offers mixed support for the expectation that women and men will differ in their views on appropriate sanctions for sexually harassing politicians. In their experimental survey, Stark and Collignon (Reference Stark and Collignon2022) do not find that American women are more willing to change their vote intentions after learning of a SH scandal. In contrast, Barnes, Beaulieu and Saxton (Reference Barnes, Beaulieu and Saxton2020) find that women are generally more likely than men to electorally punish a candidate who has been involved in any type of scandal. Compared to men, women are also more likely to support penalties for corrupt politicians (Stensöta Wängerud and Agerberg, Reference Stensöta, Wängnerud, Agerberg, Dahlstrom and Wangnerud2015). In Canadian federal politics, Raney and McGregor (Reference Raney and McGregor2023) find that women’s and men’s accountability expectations are especially likely to diverge at the time an allegation is made. They report that Canadian women are more likely than men to support the public disclosure of an allegation of SH and to support an MPs’ suspension from office pending an independent investigation. In the absence of strong partisan cues, we therefore suspect that women will be more supportive of sanctioning a politician who perpetrates SH compared to men particularly at the time an allegation is made.

Sexism & Attitudes About Sexual Harassment

At the same time, a respondent’s self-identified gender may not be the only, or perhaps even the best, predictor of attitudes about SH in politics. Together, gender role perceptions and sexism have been found to influence policy attitudes that are salient to women’s lives (Hansen and Dolan Reference Hansen2020, 533; Cassese and Barnes Reference Cassese and Barnes2018). Gender role stereotypes have been linked to perceptions of SH specifically, such that women who hold traditional gender role beliefs appear more apt than others to hold other women (and themselves) responsible when they are sexually harassed (Cesario Reference Cesario2020). Conversely, men can hold gender-egalitarian views and view SH as a serious problem. Barnes, Beaulieu and Saxton (Reference Barnes, Beaulieu and Saxton2020) find that people with hostile sexist attitudes (i.e., antipathy towards women who are viewed as taking away men’s power) are less likely than other voters to support a woman politician who is involved in a sex scandal compared to a male candidate. Those who hold benevolent sexist attitudes (i.e. a chivalrous belief in the need to protect women) were less likely than others to vote for either a woman or a man involved in a scandal. Costa et al’s (Reference Costa, Briggs, Chahal, Fried, Garg, Kriz, Milne and Slayton2020) experimental research demonstrates that hostile sexist beliefs predict electoral support for a candidate embroiled in a SH scandal. They find that sexism, rather than partisanship, is a key determinant in voter assessments of candidates who engage in SH.

Using survey data, Archer and Kam (Reference Archer and Kam2021) observe that sexist ideology correlates with views about the #MeToo movement, with those who hold “modern” sexist views comparatively more likely to minimize the pervasiveness of sexual misconduct and to believe that #MeToo has gone too far. Looking at violence directed at women politicians by citizens, Håkansson (Reference Håkansson2023b) assesses how subtle forms of sexism – by way of voters’ higher propensity to direct their grievances towards women politicians compared to their male counterparts—fuel gender-motivated political violence. Finally, using a survey experiment, Longdon and Banducci (Reference Longdon and Banducci2023) find that voters with less hostile sexist attitudes were less likely to support a candidate accused of sexual misconduct. Bringing this research together, we suspect that in addition to differences between women and men, sexist attitudes will shape voters’ opinions about legislative accountability and SH.

Beliefs About Gender-Based Violence

We are further interested in whether respondents’ opinions about gender-based violence specifically shape their accountability expectations of politicians who engages in SH. Social psychology research points to the prevalence of gender-based stereotypes and biases about violence, which refer to the “acceptance of modern myths about sexual aggression” (‘AMMSA’) (Milesi et al. Reference Milesi, Süssenbach, Bohner and Megias2019). AMMSA is defined as “descriptive or prescriptive beliefs about sexual aggression (i.e., about its scope, causes, context, and consequences) that serve to deny, downplay, or justify sexually aggressive behavior that men commit against women” (Gerger et al. Reference Gerger, Kley, Bohner and Siebler2007, 425). This research draws our attention to the ways that sexual violence myths reinforce victim-blaming mentalities, while at the same time exonerating a perpetrator from wrong-doing. Myths include how men’s sexual violence is natural and normal, that women exaggerate or falsely claim to be assaulted or harassed, feel flattered and secretly enjoy being sexually harassed, and that SH has a minimal impact on those who are targeted (Fitzgerald and Shullman; Reference Fitzgerald and Shullman1993; Milesi et al. Reference Milesi, Süssenbach, Bohner and Megias2019). Although correlated with other attitudinal variables like modern sexism, right-wing ideology, and social dominance, beliefs about gender-based violence arguably constitute a unique form of moral reasoning that might influence perceptions of perpetrator accountability. By casting doubt on the “believability” of women’s claims of violence, adherence to such myths helps to cognitively deflect blame away from perpetrators, thereby reducing or eliminating the need for accountability altogether.

Attitudes about gender-based violence could also play out differently for women and men, reinforcing any observed gender-based differences in our study. Lonsway, Cortina and Magley (Reference Lonsway, Cortina and Magley2008: 612) find that men are more likely than women to accept gender-based violence myths, while women are reluctant to report their experiences of sexual violence to the authorities for multiple reasons, including the knowledge that they may not be believed (Gilmore Reference Gilmore2023). The #MeToo movement helped expose the gendered and raced nature of this “credibility discount gap” where historically, women and girls (especially Black women and girls) targeted by sexual violence have not been believed, while (white) men perpetrators have evaded accountability (Burke Reference Burke2021; Gilmore Reference Gilmore2023, 45). We suspect that respondents’ beliefs about gender-based violence will have an important and unique effect on their support for sanctioning politicians who engage in this behavior, and that such opinions might vary for women and men.

Who Supports Punishment?

Next, we consider in more detail the correlates of support for punishment. To do so, we create two indices that indicate support for punishment in the two conditions – allegation and finding. These measures reflect the number of punishments that respondents would support (of six) in each condition. These variables range from zero to six, with high values indicating support for a higher number of punishments in the “findings” condition. As explanatory variables, we also include several sociodemographic and attitudinal factors: gender, age, education, income, relationship status, employment status, place of birth, race, and religion.Footnote ¹⁶ As an additional control, we also include a measure of political ideology.

To measure attitudes about sexism, we create an index of two agree/disagree questions that ask respondents their opinions of whether sexism is a problem in the workplace and whether #MeToo has been positive for society, with the general expectation that those who score high/low on this measure will be comparatively supportive of/resistant to punishment (as adapted from Cossette and Craig Reference Cossette and Craig2020, 85-86). The next variable is our attitudinal measure of gender-based violence, which includes whether a respondent believes that: (1) women often mistakenly interpret innocent remarks or actions by men as sexism and (2) there are too many people who are claiming SH or assault. Prior studies have used these questions separately to test for generalized sexism in U.S. politics (Cossette and Craig Reference Cossette and Craig2020, 82-84). Here, we use them to create a unique composite index related to gendered myths that women are prone to exaggerate and make false claims of their experiences of sexual violence, and we label this variable “skepticism.”Footnote ¹⁷ We then test the potential of this “skepticism” index to predict voters’ support for sanctioning a politician who engages in SH.Footnote ¹⁸ Finally, we use an indicator of whether respondents have personally experienced unwanted sexual attention or harassment, under the assumption that these experiences might make respondents supportive of punishment for perpetrators of SH.

We run OLS models to identify the correlates of support for punishment in both the “allegation” and “finding” conditions. This approach allows us to assess whether the same factors are associated with attitudes towards punishment following an allegation and after a finding. Raney and McGregor (Reference Raney and McGregor2023) employ a similar test, but ours is an improvement in that we ask about the same punishments at both the allegation and finding stages; whereas in their work, they consider different punishments at each stage of a claim. We also consider six types of punishment at each stage, while they consider only two.

What are the correlates of attitudes towards punishment? Figure 1 shows coefficient plots for two OLS models, where the “allegation” and “finding” indices serve as the outcome variables. Recall that these variables range from 0 (indicating support for none of the punishments) to 6 (support for all punishments). The socio-demographic variables are all dummies, as is the “personal experience” variable.Footnote ¹⁹ Ideology, as well as the “sexism is a problem” and “skepticism” variables are interval level. All explanatory variables are coded to range from 0 (minimum) to 1 (maximum). The figure shows point estimates and whiskers indicate 95% confidence intervals. Black results are significant at the 95% confidence level, while grey results are not. Full model results are found in Appendix III.

Figure 1. Correlates of support for punishment: Allegation and Finding.

Figure 1 shows the factors that explain respondents’ attitudes towards punishment. In the “allegation” model, several sociodemographic factors are significant, with women, individuals without a university education, those who are married/common law, immigrants, and Catholics comparatively supportive of punishment. Our two main attitudinal variables of interest are also statistically significant. On average, respondents who believe that sexism is a problem are very supportive of punishment – this is, by far, the largest effect in the figure. Confirming our research expectation, individuals who are highly skeptical of gender-based violence claims are less supportive of punishment. The “women” variable deserves some particular attention here, given the gender-based nature of SH. Although it is related to many of the other factors in the model (including all the “attitudinal” variables), it nevertheless retains an independent effect upon attitudes towards punishment at the allegation phase.Footnote ²⁰ Women and men may have different views on whether sexism is a problem or on gender-based violence but even after controlling for such factors, self-identified gender still matters.

The results of the “finding” model are slightly different than the “allegation” results, with fewer of the sociodemographic characteristics reaching statistical significance. Notably, we observe that the gender gap has disappeared, while age is the only statistically significant sociodemographic factor (with older individuals being more supportive of punishment). Right-wing voters appear less supportive of punishment compared to left-wing voters – a gap that did not appear in the “allegation” model. The “personal experience” variable is insignificant in both models. Finally, both the “sexism” and “skepticism” variables remain significant, and in the same direction as the previous model.

A particularly noteworthy feature of Figure 1 is the differences between the two models – some variables are significant at the allegation stage, but not after a finding (and vice-versa). To determine if these differences are statistically significant, we run a third model, where the outcome variable is the difference between the number of punishments respondents support after a finding and after an allegation. Values for this new variable range from –6 to 6, with a high value indicating support for more punishments in the “finding” condition. Figure 2 shows the results of another OLS model with this new variable as the dependent variable, using the same explanatory factors as in Figure 1. Statistically significant results here indicate that the explanatory variables are differently related to attitudes towards punishment in the two points in time. As above, the figure shows point estimates and 95% confidence intervals.

Figure 2. Correlates of support for punishment: Finding minus Allegation.

Figure 2 shows that a number of factors are differently related to support for punishment at the finding and allegation stages. This setup is important as it shows how support for punishment changes across the two stages. The negative result for “women” confirms what we observed in the two separate models in Figure 1: the gap between women and men that we find in the allegations model is, in fact, different from the null result for that variable observed in the “finding” model. Some other sociodemographic indicators are also significant, with a gap between older and younger respondents, university educated and less than university educated respondents, high-income and less than high-income voters, and Catholics and Atheists. The model in Figure 2 tells us that the effect of these factors on support for punishment of a councillor depends on whether an allegation of SH has been substantiated through a formal investigation.

Turning to the attitudinal variables, both the significant and insignificant results are of interest. The “sexism is a problem” variable is significant in all models in Figures 1 and 2. The negative result in Figure 2, however, suggests that this factor matters less for attitudes towards punishment in the “finding” condition. Interestingly, we also find that the “skepticism” variable is significant in both models in Figure 1, but is insignificant here, meaning that this factor matters the same for views on punishment in both conditions. This result suggests that skepticism about gender-based violence is at play no matter the stage of a SH claim, or put another way: a skeptic is always a skeptic. We also find that ideology is significant (and negative), which accords with the finding from Figure 1 that left- and right-wingers differ from one another at the “findings” stage only.Footnote ²¹

All in all, the results in Figure 2 indicate that the question of whether an allegation of SH is substantiated by an investigation matters to respondents, and that it matters quite a lot. Support for punishing a councillor not only increases after a finding of SH (as we see in Table 1), the correlates of support also change at this stage of a claim. Of note is again the difference between women and men respondents: while the importance of other factors changes following an SH finding, the diminishment of a gendered attitudinal gap is noteworthy.Footnote ²²

Does the Type of Punishment Matter?

Despite the observations in Figures 1 and 2 that show no substantive differences between women and men at the findings stage, we are further interested in whether women and men might support different types of punishment depending on their severity. The outcome variables in Figure 1 and 2 are based on indices, which are counts of support for up to six punishments. In this final stage of our analysis, we consider the correlates of support for each of the six individual punishments following a finding of SH.

To test whether the self-identified gender of a respondent matters for specific sanctions, we run a series of six different logistic regression models. The explanatory variables considered are the same as those included above, but the outcome variables are now binary indicators of whether respondents would support (or not) each punishment following a finding of SH (where the punishment index suggests that there is no difference between women and men). We present the results for this final step of our analysis in Figure 3, which shows the findings for the gender variable for each of the six models – one for each punishment (listed on the left side of the figure). The coefficient plots show the marginal effect of gender in each “punishment,” with 95% confidence intervals, along the x-axis. Punishments range from gentlest (at the top of the figure) to harshest (at the bottom). The full model results are found in Appendix III.

Figure 3. Support for punishment following a sexual harassment finding.

In Figure 3, an estimate above 0 indicates that women are more likely men to support a particular form of punishment. Again, black estimates are statistically significant at P = 0.05. Recall that in Figure 1, a respondent’s self-identified gender appeared to exert no independent effect on attitudes towards punishment at the aggregate-level following an SH finding. In Figure 3, we disaggregate support for punishment at the findings stage, where important differences between women and men emerge based on the severity of the punishment. For the gentler or more moderate sanctions – making the public aware of the misconduct, forcing the councillor to take training or give an apology, and deducting their pay – women and men are no different from one another. However, for the two most severe punishments that involve removing the councillor from office, a gender gap in opinions reappears. As demonstrated in Figure 3, women are more likely than men to believe a councillor found to have committed SH should be removed from office (at the 95% confidence level) and to support banning them from running for office again (P = 0.06). To be clear, both women and men are highly supportive of removing councillors from office at this point – as we saw in Table 1. Women are, however, roughly five percentage points more likely than men to believe that a councillor should be removed from office after a finding of SH. In short, the presence of a gender gap in attitudes towards the punishment of councillors found to have committed misconduct depends upon the type of punishment being imposed.