Moral licence and disbelief: how voters look past political misconduct

ABSTRACT

How do voters respond to a co-partisan political candidate after hearing about a serious scandal? We apply qualitative content analysis to open-ended text responses collected from survey respondents, focusing on 159 respondents who remain willing to vote for a hypothetical male candidate after learning that allegations of sexual harassment against him were settled out of court. We uncover two main strategies by which respondents explained and justified their willingness to look past the misconduct allegations. The dominant strategy was to extend moral licence to the candidate based on his prior good deeds; a second prominent strategy was to disbelieve and reject the allegations. Our findings offer new insights on why some political candidates and careers appear untroubled by even serious allegations of misconduct, and frames new research avenues on which voters might be more likely to extend moral licence and in what circumstances.

KEYWORDS:

• Voting behaviour
• political misconduct
• moral licence
• disbelief
• thematic analysis

Introduction

Politicians being accused of sexual misconduct is neither a novel nor a rare phenomenon. In the US, election campaigns at all levels have aired allegations of sexual misconduct, and Trump became the first ex-president to be found guilty of sexual harassment and abuse by a court of law in 2023.¹ He remained the Republican party’s frontrunner in polls for Primary elections and went on to secure the party’s nomination; and retains considerable support for the general election even while sitting in live court proceedings for a separate court case about financial misconduct.² This phenomenon is not exclusive to the US. Reporting of recent allegations of sexual misconduct in the UK was dubbed a ‘Westminster Too’ moment (Krook 2018). Even now cases continue to be reported, prompting a number of high-profile by-elections in the UK over 2023 and 2024.

Investigating citizen responses to news of political misconduct is important to our understanding of voting behaviour; the reasoning voters use to evaluate candidates; and to broader questions of fairness by which forgiveness appears to be extended to some candidates in some situations, and not others. Much of the existing scholarship on political scandal generally, and sexual misconduct cases specifically, aims to establish and quantify the electoral penalties that voters exact on accused candidates. Past findings show that while misconduct can lead to adverse electoral outcomes, such consequences are not consistent. They may depend on electoral context, candidate traits and voter attitudes, partisanship and personal values among other factors.

The converse question of when electoral penalties appear not to apply is equally important, and may help explain why real world electoral outcomes sometimes fail to replicate the sizeable and negative effects we would expect from survey experiment findings. Some literature indicates that voters forget – the sizeable informational effects early in a campaign have been shown to dwindle as time passes (Mitchell 2014; Vonnahme 2014). And voters appear to forgive – particularly when the allegations are against an otherwise-liked leader or co-partisan, or in close election competitions (Lee et al. 2023). Research has increasingly looked at why voters appear to tolerate corruption; however, we know less about voters who do not penalize politicians accused of sexual misconduct.

In this article we explore the strategies that voters draw upon to look past allegations of misconduct, and remain favourable towards an accused political candidate. We ask: how do voters explain their willingness to vote for a candidate after learning of sexual harassment allegations? We qualitatively analyse themes emerging from text responses drawn from an online survey of US voters, with a deliberate exploratory approach to discover new insights. We focus our attention on a sub-sample of 159 respondents who were informed of sexual misconduct allegations against a hypothetical co-partisan candidate, and who then indicated positive voting intentions for that candidate. Our findings uncover support for two significant strategies. The first is moral licensing, where ‘past good deeds can liberate individuals to engage in behaviours that are immoral, unethical, or otherwise problematic' (Merritt, Effron, and Monin 2010, 344). Around 45% of our voters who remain favourable towards the candidate indicate a form of moral licensing: they cite perceived positive traits and track record of the accused candidate, and use these positive aspects to outweigh or downplay the allegations. The second strategy is disbelief, with just over a quarter of voters in our sample simply not believing the allegations and dismissing them as ‘fake news'. The latter finding supports other studies that identify disbelief as playing a role in the evaluation of sexual abuse allegations and victims (Harmer and Lewis 2022).

The moral licensing strategy has not, to the best of our knowledge, previously been linked to explanations of voter behaviour. Our evidence suggests moral licence can be lent to co-partisans as a way of acknowledging and then looking past bad actions. We note that ‘looking past’ is not the same as forgiving a candidate, rather the respondent appears not to have reached the threshold for electoral punishment even if they dislike or disapprove of the allegations. We undertake brief statistical analysis to investigate who among our respondents are more likely to apply the moral licensing strategy, and find indicative evidence that marital status, trust in people, and partisan identity are associated with using strategy.

Our study frames new questions for the study of political misconduct going forward. Firstly, we demonstrate the value of qualitative analysis to unpack and understand voter responses to scandal. Secondly, the findings around moral licence offer fresh insights on how voters might look past other forms of bad political behaviour – corruption, anti-democratic rhetoric and actions, backsliding on human rights, for example – and remain supportive of political candidacies. We speculate that our mechanisms are two (amongst other) observable patterns of behaviour which partisan motivated reasoning might give rise to; and can potentially enhance our understanding of candidate evaluation, accountability and updating views in response to new information.

The article proceeds as follows. Section 2 reviews prior research. Section 3 summarizes our research design, data and thematic analysis approach. Section 4 presents our findings; and Section 5 concludes.

Misconduct and voter responses

Prior research offers a broad consensus that political figures are negatively affected by scandals (Banducci and Karp 1994; Vonnahme 2014; Welch and Hibbing 1997). Some estimates suggest that a scandal can subtract almost 4% of the incumbent’s total vote, with the candidate 13 times more likely to lose than a scandal-free incumbent (Basinger 2012). There is also a negative effect on the candidate’s decision to stand for re-election (Banducci and Karp 1994), with some opting to resign or retire (Basinger 2012).

Robust empirical evidence, and anecdotal observation, indicates that not all candidates see their political careers affected to the same degree by scandals. Why do effects vary? One source of variation relates to the type of scandal: whether it is financial, corruption, political or sexual, for example (Barnes, Beaulieu, and Saxton 2020; Basinger 2012; Doherty, Dowling, and Miller 2011). In corruption cases, if the scandal generates side benefits for voters they are less likely to punish a corrupt candidate (Fernández-Vázquez, Barberá, and Rivero 2016). Another factor relates to the political candidate’s demographic characteristics, including race (Berinsky et al. 2011), gender (Bhatti, Hansen, and Leth Olsen 2013), and sexuality (Rajan and Pao 2022); and their perceived personality traits such as empathy (Funk 1996). The electoral context matters too, such as the availability of a ‘clean alternative’ candidate to vote for (Agerberg 2020; Jankowski, Juen, and Lewandowsky 2023).

A candidate’s use of blame avoidance strategies can be effective in minimising voters’ negative evaluations, particularly for co-partisans and among left wing voters (Schönhage and Geys 2024). Negative effects of scandalous information tend to reduce over time, which may explain why some politicians appear not to face any penalties at the ballot box (Mitchell 2014; Vonnahme 2014). The hearings of Supreme Court Justices Clarence Thomas in 1991 and Brett Kavanuagh in 2018 prompted analysis of how public opinion moved as evidence unfolded, and found that opinion often came to rest in favour of the accused men (Mansbridge and Tate 1992; Frankovic and Gelb 1992; Thomas, McCoy, and McBride 1993; Herrera Hernandez and Oswald 2022).

A further explanation for why the effects of misconduct vary relates to voters themselves: their characteristics, attitudes and the cues they use to evaluate candidates. Recent research points to the role played by voter attitudes towards women and the prevalence of hostile and benevolent sexism among those attitudes (Barnes, Beaulieu, and Saxton 2020; Herrera Hernandez and Oswald 2022); as well as voters’s political ideology (Bhatti, Hansen, and Leth Olsen 2013; Saxton and Barnes 2022) and party affiliation (Stark and Collignon 2022). Personal values have also been shown to matter, particularly self-transcendent values – universalism and benevolence (Schwartz 2006) – which reflect how much emphasis one places on the welfare and wellbeing of others (Savani and Collignon 2023).

Existing literature underscores the importance of party loyalty as a reason for people overlooking serious allegations of misconduct and corruption (Lee et al. 2023; Hansen and Dolan 2020; Muñoz, Anduiza, and Gallego 2016; Cossette and Craig 2020; Jankowski, Juen, and Lewandowsky 2023). But here too can be shades of nuance. People vary in the way they evaluate accused co-partisans (Costa et al. 2020), suggesting that the party heuristic cannot explain the entirety of voter reasoning. There may be other complex psychological processes involved, which require voters to resolve the dissonance involved in evaluating a preferred candidate accused of bad behaviour. We aim to investigate further the apparent reasons for a lack of negative response to news that a co-partisan candidate was accused of sexual harassment.

Research design

Much of the scholarship mentioned here relies on survey experiments and quantitative outcomes relating to voting intention and voter perceptions, measured on attitudinal scales for traits such as trustworthiness and favourability (Rajan and Pao 2022). Where we differ in this article is by adopting a qualitative approach to understand voter reasoning. We analyse voters’ reflections in their own words and expression, identify and organize these reflections into themes, and leverage a richer understanding of how they arrive at and explain their voting intention. This is not an entirely new approach: Mansbridge and Tate (1992) draw on interviews with Black women to support their analysis of public reactions to Anita Hill’s testimony against Clarence Thomas. A small number of studies have opted for in-depth qualitative analysis of voter attitudes towards broader political transgressions. For example, Mathur (2023) draws on interviews with Indian voters to investigate how they resolve dissonance around seeing bad news stories associated with a highly popular Prime Minister Modi. Gaskell, Stoker, and Jennings (2023) investigate voters’ attitudes towards political leaders where a failure of trust has occurred, using focus group discussions to understand the reasons for forgiveness being withheld or granted. Our study focuses on open text survey responses from voters who are not minded to penalize a candidate accused of sexual harassment. How do they make sense of the allegations, and how do they move past them? How do they explain the candidate’s favourability after learning about the alleged misconduct?

Our approach based on the analysis of the open-ended text answers of questions presented during a survey experiment is well suited to address these questions and presents at least three advantages over other methodologies: (a) it allows us to explore voter’s reasoning in their own words, providing us a deeper understanding of why some voters remained loyal to the accused candidate relative to closed survey questions; (b) it reduces the risk of social desirability bias, given the nature of the informational vignettes where respondents might have felt social pressure to provide a normatively or politically ‘correct’ response; (c) it offers the possibility of opening the ‘black box’ of informational treatment effects (Hesse-Biber 2013; Tarrow 2010) to enrich our understanding of voter reasoning.

Our survey data is drawn from a broader study which implemented an online survey to 704 participants (see Appendix Table A1 for descriptive statistics and A2 for the survey instrument).³ The survey offered an introductory vignette about a hypothetical male candidate running for office as Governor, including his policy priorities, track record in politics, and credentials as a family man who cares for the community. The next stage of the survey randomly selected approximately half the participants to receive information of media reporting that a lawsuit had been settled after allegations of sexual misconduct by the candidate’s staffers (see Table 1). The other half of the sample, a control group, received an alternate informational vignette about campaign trail activities with no information relating to a lawsuit or allegations of misconduct. All respondents were then asked to rate their voting intention using a four-point categorical scale (definitely not/likely not/likely yes/ definitely yes).

Table 1. Information provided to survey respondents.

Vignette 1 (introduction – seen by all participants)	David Anderson is your preferred party’s candidate for State Governor. Anderson previously served as a U.S. Representative for 10 years, and has a strong record on job creation and working to minimise the country’s deficit. Anderson supports business and urban development schemes and has promised to help all citizens ‘achieve their full potential'. As U.S. Representative, Anderson developed a reputation for a great work ethic, strong ties to the community, and backing local services. Anderson and his wife moved to the state twenty years ago; the couple raised two children together, both of whom now attend universities in the state.
Vignette 2a (no misconduct allegations – seen by control group)	Recently there has been some media coverage on candidate David Anderson. He visited primary schools and launched his campaign promise to improve the quality of education by investing in teachers and pupils.
Vignette 2b (sexual misconduct allegations – seen by treatment group)	Recently there has been some media coverage on candidate David Anderson. Two former staffers went public with the accusation that Anderson had sexually harassed them three years ago. It was revealed the parties settled a lawsuit about the matter.

As reported separately, both groups reported a high degree of support for the candidate following an introductory vignette (over 90% were definitely or likely to vote for the candidate), hearing about the scandal went on to reduce the probability of voting for the candidate by 51% (Savani and Collignon 2023). In other words, the misconduct led to a sizeable and statistically significant negative effect on voting intention. However, a substantial minority of the participants – 159 respondents, or 43% of those who received information about the misconduct – remained open to voting for the accused candidate. The heterogeneity is partly explained by personal values, drawing on Schwartz (2006), particularly those relating to self-transcendence. Voters who strongly identified with values of universalism and benevolence were less likely to vote for the candidate after learning of the allegations (see Savani and Collignon (2023) for full results from the experimental component of the study).

The survey concluded with an open-ended text response box where respondents could explain their voting intention. These responses make it possible to investigate, in the participants’ own words, what made some of them not switch away from the accused candidate. We report below the process and findings from this inductive and exploratory analysis, bearing in mind that the open text response may be an opportunity to reflect and explain, but also to defend and motivate ex post.

We apply a data-driven qualitative content analysis approach (Hsieh and Shannon 2005) using manual coding of open-text responses from 159 respondents, selected on the basis they had (1) heard about the misconduct allegations and (2) were ‘likely’ or ‘definitely likely’ to vote for the accused candidate. The qualitative content analysis involved three coders who initially looked at the full dataset of text responses.⁴ Two coders conducted a preliminary analysis using a random sub-sample of 50 respondents. We identified a series of broad themes independently, which were then discussed and used to agree upon a preliminary coding scheme. Coding used sentences as individual units of analysis, and in some cases, a unit may have been coded to more than one theme and would count for both. The next stage involved two coders (one of the original coders, one new) applying the preliminary coding scheme to the full dataset (using NVivo version 12). This led to further discussion and iteration of the themes (Cheung and Tai 2023), and some reconciliation of coding (such that the researchers agreed on 99.97% coded items of text), allowing us to agree upon the emergent themes and prioritize them by the frequency with which they arose.

Results

Seven distinct themes emerge from voters who know of the accusations and are positively inclined to vote for the co-partisan candidate (see Table 2). The two most prominent of these indicate (1) a degree of behavioural licence for the candidate, and (2) a disbelief of the allegations. The nature of the coding means that one respondent can offer more than explanation for their voting decision, and in a few cases, there is overlap across the themes. This is a plausible reflection of real-world reasoning. We do not try to separate out respondents to place them in only one category (moral licensers) or the other (disbelievers); rather, we see it as an additional insight that we can capture the small degree of overlap between different mechanisms.⁵

Table 2. Evaluations from voters who learn of the allegations and support the candidate.

	Theme: Moral Licence	Excerpts of coded references
1	Draws on other positive factors to balance out the allegations, wants to give a second chance 57 respondents mentioned this theme (36% of sample) 33% of text was coded to this theme	‘The positive aspect of what he has done before' (id 318) ‘His work history and ability to get things done were most important' (id 253) ‘He wants to fix the budget and has 10 years of political experience' (id 107) ‘I think what he has done for the state in the past should speak loudly' (id 163) ‘If really he done that mistake we can give a chance to accept his mistake because he is good at his work for [the] past ten years' (id 548)
2	Policy platform viewed favourably 19 respondents mentioned this theme (12% of sample) 15% of text was coded to this theme	‘His policies and practices align with me' (id 71) ‘Experience; fiscal responsibility; job development' (id 159) ‘To me, his plans for the state as governor were most important to me over harassment accusations. I like the vision he has for how he would lead the state’. (id 50)
3	Relevance of scandal to office 10 respondents mentioned this theme (6% of sample) 9% of text was coded to this theme	‘I believe the candidate's professional record in politics and work life are the most important factors that cause me to evaluate him positively' (id 664) ‘I firmly believe a politician does not have to be a moral person in their personal life to be a competent leader, so the harassment allegations and settlement did not sway me too much' (id 50)
	Theme: Disbelief	Excerpts of coded references
4	Lack of trust in allegations, including due to time elapsed 30 respondents mentioned this theme (19% of sample) 28% of text was coded to this theme	‘They’re just trying to extort money, should have come forward at the time’ (id 56)
	Theme: Other	Excerpts of coded references
5	Political scandal seen as normal, voter desensitized to allegations 11% of text was coded to this theme	‘Seems like an everyday thing with politics. People accusing each other' (id 132)
6	Want to know more before punishing candidate 11% of text was coded to this theme	‘The nature of the allegation and the settlement does not seem too alarming in my personal opinion. He has a strong work history and I would need more information about this to feel differently' (id 181)
7	Party identification and partisanship 8% of text was coded to this theme	‘I'm assuming that it's a battle between this guy who sounds like a Democrat and a criminal Republican. In this case, I have no choice but to avoid supporting the guy who will add to my problems i.e. the Republican' (id 11)

Notes: Individual respondents may have been coded to more than one theme so percentages do not add up to 100%.

As Table 2 shows, the moral licensing strategy was found to operate in three distinct but related ways, and we therefore capture moral licence in a broader sense using sub-themes 1–3. The first tends to relate the voter’s favourable response to the candidate’s track record and previous good deeds. The second and third are grounded in the present more than the past, for example where a policy position is used as a positive frame to outweigh or take precedence over the allegations (theme 2); or where the allegations are diminished in their relative importance compared to the candidate’s other qualities such as leadership or political competence (theme 3). The common thread is that positive information and attributes of the candidate are used in the voter’s reasoning to look beyond the misconduct allegations. Examples of coded text are set out in Table 2.

Moral licensing

Moral licensing implies that past behaviours or achievements are used as the basis of permission for less virtuous behaviours now or in the future. Anecdotal examples in everyday life abound – for example, I went on a morning run, I deserve that afternoon donut. In one of a small number of studies that consider moral licensing not of one’s own behaviours but of someone else’s, Effron and Monin explain that ‘licencing occurs when observers reduce their condemnation of morally dubious behaviour or the actor who committed it in light of the actor’s prior good deeds' (2010, 1618). In our data, aspects of prior good deeds and other virtues are used to counteract and excuse the less desirable behaviour implied in the scandal vignette – precisely the definition of giving moral licence.

The most frequently recurring theme saw voters justifying their favourable vote by weighing up other positive features of the candidate against the negative allegations (theme 1). A number of respondents returned to ‘the positive aspect[s] of what he has done before' (id 318) as a reason for why the candidate was seen in a positive light after the initial vignette. Respondents liked the idea of his having held public office before, referring for example to his ‘experience in politics and serving the people' (id 435), and the idea that ‘he wants to fix the budget and has 10 years of political experience' (id 107), and ‘his past record of job creation and developing urban areas, having a good work ethic' (id 232).

Another response pointed to ‘his ties to the community, his work ethic, his status as a family man, his ability to withstand adversity' (id 607). In this excerpt, it is interesting to see language around ‘adversity’, which was not explicitly mentioned in vignette 1. It may suggest voters who form a positive first impression embellish those positive characteristics; or the respondent may have been referring to the allegations as the source of the adversity. These comments clearly indicate that the respondent’s sympathy lay with the candidate.

A prevailing idea within this theme was the candidate should be viewed in the round, with the allegations being just one amongst many factors to consider and weigh up:

I think that regardless of who you choose, there is going to be some sort of controversy, so I still want to focus on the candidate as a whole (id 484)

The allegations were sometimes dismissed as a mistake, suggesting the lawsuit might be an anomaly in an otherwise positive track record, which should not prevent the candidate from being voted in as Governor. Closely linked to this is the idea of deserving a second chance. For some respondents, the allegations appear to be perceived as neither career-defining nor career-ending:

That might be a mistake and doesn't necessarily describe the type of person he is (id 602)

His past experience in politics and the family life he leads. He also supports community and local businesses. Everyone makes mistakes (id 256)

Other strengths of the candidate such as his policy positions were taken as positive traits that were more relevant than the allegations (theme 2), including a perceived fiscally conservative approach to debt and public finances, as well as job creation. One respondent explicitly prioritized the ‘economy and the deficit. I’ll reserve judgement on the accusations' (id 716). Despite the introductory vignette being so brief, respondents were able to relate moderately strongly with the candidate’s political vision, with one respondent stating: ‘His policies and practices align with me’ (id 71), and another praising the candidate’s ‘Experience; fiscal responsibility; job development' (id 159).

For some respondents the allegations were viewed as less important to the requirements of office than a professional track record (theme 3). For example, some excerpts highlighted aspects of the candidate’s policy platform that were important and outweighed the personal allegations:

I wouldn't seek his friendship, his personality traits need improvement, but his political experience seems adequate (id 296)

These themes lend weight to the idea that voters were giving moral licence to the candidate. Taken together, themes 1, 2, and 3 cover a significant share of the responses (72 respondents in total), suggesting a prominent role for moral licensing expressed in these different forms. It is both possible and plausible that this is related to the candidate being presented to our survey respondents as a co-partisan. Our findings relate well to those of Munoz et al (2016, 603–604), who identify ‘micro-mechanisms’ in voter evaluations of corruption allegations.

Disbelief and rejection of the allegations

As Table 2 demonstrates, however, moral licensing is not the only mechanism at work. The second most frequently coded theme related to trust in the reports (theme 4, with 30 respondents coded). A number of respondents were concerned about the veracity of the allegations, and the potential for ‘filing false claims', and making ‘false accusation'. Overall, our data points to there being a significant group of voters who simply do not believe the allegations, and put forward various ways to justify their rejection of the allegations (theme 4).

One justification relates to the timing of the news story in relation to the alleged events (described as having taken place three years in the past).

I've grown tired of these sexual harassment suits that come up after years of silence. Someone run[s] for a political office and suddenly these people come out of the woodwork, it does not affect my vote anymore. (id 688)

It’s fake news, why did it take the staffers 3 years to report it (id 357)

These responses echo the findings of earlier studies. Mansbridge and Tate (1992) report from their interviews that some women found it problematic that Anita Hill had waited to come forward with allegations of sexual harassment, an idea echoed by Frankovic and Gelb (1992) in their analysis of public opinion more generally.

There is also a readiness to ascribe political motivations to the allegations in order to reject them:

Candidates for such an office are generally targeted with claims such as this and without further details to attempt to find out if the allegations are true and what the extent of claimed impropriety is in fact. (id 482)

… those that came out publicly now didn't come out before his aspiration to be the Governor (id 85)

Some people use the excuse of sexual accusation to discredit politicians (id 535)

Here too we find parallels in prior research, such as a recent finding that support among co-partisans for Brett Kavanaugh’s confirmation was ‘not perceived as entirely about sexual harassment but increasingly as political issues' (Herrera Hernandez and Oswald 2022, 1381).

The increase in reports of sexual harassment in politics and more widely appears to have produced a backlash among this group of respondents. The excerpt below implies the accusers are jumping on a bandwagon, and once again the timeframe of the allegations is viewed with suspicion.

… Because of MeToo, information from the past is getting a bit hazy and suspicious for certain candidates (id 79)

For a sizeable proportion of voters who withdrew their support for the candidate, the vignette’s reference to a lawsuit being settled was significant. Some 22% of respondents who did not support the candidate after hearing of the allegations explicitly mention the lawsuit as a potential sign of guilt or having something to hide (see Appendix Table A2). In contrast, the legal process was far less influential to the voting intention of those who reject the allegations, with two respondents explaining:

Lawsuits are routinely settled with the defendant paying just to get rid of the lawsuit, and is not an admission of guilt (id 10)

It's not just a matter of being accused and settling out of court. that doesn't imply anything at all except to stupid people. I would require knowing who the accusers are and their reputation and motivation for lying before I would make such a choice (id 339)

In each of these cases, the disbelief of the allegations is explained and defended, based on the reporting being misleading, or the allegations false and either financially or politically motivated. The use of disbelief and skepticism over the nature of evidence provided has also been recorded in online discussions of sexual harassment against women (Harmer and Lewis 2022). The responses we uncover in defence of preferred political candidates have also been found in studies of corruption (Anduiza, Gallego, and Muñoz 2013; Cornejo 2023) and moral violations (Walter and Redlawsk 2023).

Other reasons for remaining favourable towards the candidate

Other themes pick up a view that the allegations are a form of ‘business as usual', to the extent that the respondents appear somewhat inured to questions of serious misconduct:

These days accusations are a dime a dozen (id 525)

The nature of the allegation and the settlement does not seem too alarming in my personal opinion (id 181)

To a lesser extent, some voting decisions were framed in terms of not wanting to advantage the competing party (8 such references in total). But on the whole, voters that were favourable to the candidate after reading the treatment vignette were more likely to react to the detail of the allegations rather than explicitly fall back on co-partisan loyalty.

Who relies on moral licence as a strategy to look past misconduct?

We now delve deeper into our survey dataset to see if we can identify any political, demographic or attitudinal traits that are associated with using moral licence as a way of moving past the allegations. We undertake further statistical analysis on the moral licenser group which is a larger sub-sample. The ‘moral licensers’ are those respondents who were coded positively for the moral licence strategy (themes 1–3). From our qualitative analysis we identified 72 moral licensers in total and 30 disbelievers (coded positively for theme 4), of whom 8 also gave indication of moral licensing (and are included in the group of moral licensers below). Caution is required with interpretation of our descriptive statistics in Table 3, but it is nevertheless possible to compare the sub-groups (moral licensers and non-moral licensers) for any defining characteristics.

Table 3. Of those who look past the misconduct, who are the moral licensers?

Characteristics of treatment group members with positive voting intention (n = 159)	Moral licensers (n = 72)	Non-moral licensers (n = 87)	Difference (p-value)
Age (modal)	30s	30s	0.234
Female (%)	0.40	0.38	0.763
Married (%)	0.57	0.79	0.002
Education (modal)	UG degree	UG degree	0.074
Holds politically liberal views (%)	0.47	0.37	0.184
Political knowledge (%)	0.76	0.71	0.466
Trusts in people in general (%)	0.56	0.72	0.027
Trusts in media (%)	0.71	0.66	0.475
Strong or very strong Republican partisan identity (%)	0.40	0.43	0.774
Strong or very strong Democrat partisan identity (%)	0.44	0.25	0.011
Self transcendence values score	3.72	2.94	0.000
Self enhancement values score	3.91	3.73	0.220

Notes: mean values or percentages displayed for each characteristic. Hypotheses tested with t-tests values scores, pr-tests for all binary variables, and Wilcoxon tests for categorical variables.

We consider a range of traits from our survey data spanning demographic and political characteristics, and personal values. Demographic factors include age, sex, educational background, and marital status. We measure political attitudes using variables that indicate political knowledge, whether the respondent considers themselves to have politically liberal views (rather than politically conservative views), whether they generally trust in people and in the media, and the strength of their partisan identity. We also ask a question to gauge political knowledge (the respondent scores 1 if they answered correctly). Previous literature has suggested partisan identity is important predictor of whether voters tend to punish or forgive accusations (Stark and Collignon 2022), and here we are able to appraise whether partisan identity and/or political views are associated with moral licensing.

We also examine associations with personal values, measured using Schwartz’s framework (2006), with a focus on self-transcendent and self-enhancement values. Self-transcendence refers to the values of benevolence and universalism; self-enhancement refers to power, achievement and hedonism. A higher score on these scales indicates higher priority is given to those values. In the context of sexual misconduct, self-transcendent values help explain how voters respond, and tend to predict a voter’s electoral punishment of an accused candidate (Savani and Collignon 2023). Self-enhancement, on the other hand, places a greater emphasis on one’s own improvement, ambition and power.

We find no clear differences in moral licensing behaviour based on age, sex, or education; but moral licensers are less likely to be married (57% compared to 79% of the non-moral licensers, p = 0.002). Among the political variables, political knowledge and holding politically liberal views do not differ, and neither does the level of trust in the media to any significant degree.

Some clear differences emerge based on other characteristics. Firstly, only 56% of moral licensers say they trust people in general, compared to 72% of non-moral licensers (p = 0.027). Secondly, the strength of Democrat party identification matters, with 44% of moral licensers reporting a strong or very strong identification with the Democrat party compared to 25% of non moral-licensers (p = 0.011). There is no difference among individuals identifying strongly with the Republican party. And thirdly, the strength of self-transcendence values appears also to differentiate moral licensers (p = 0.000). This, however, is a little surprising, as prior research has shown that holding self-transcendent values is associated with punishing the candidate (Savani and Collignon 2023). In contrast our data suggests holding self-transcendent values does not preclude strategies that appear to excuse the allegations, and suggests that further research is required to resolve the potential dissonance between emphasizing the welfare of others and finding ways to look past serious misconduct allegations. We speculate that there may be a further distinction amongst voters who report higher self-transcendence values: between those who prioritize the welfare of the staffers who raise the allegations (which would be consistent with not supporting the candidate) and those voters who prioritize the welfare of the accused candidate (which would be consistent with finding reasons to remain favourable). Potential variation in whether a voter identifies an accuser or the accused as the perceived ‘victim’ of a misconduct scenario is something that could usefully be explored in future research, and may help explain how moral licensing strategies are consistent with holding personal values of benevolence and universalism.

Limitations

Our survey was completed by AMT respondents, and the quality of text responses does vary within the sample. However, with the use of attention checks and by sifting out low-quality responses we are confident that the qualitative data offers meaningful insights into voters’ evaluations of the sexual harassment allegations and the candidate as a whole. We do not claim to offer causal evidence based on the qualitative research design; rather our findings pave the way for further investigation. It is feasible to test with larger samples in future the impact of moral licensing on voter responses to political candidates’ behaviour, and the evidence presented here suggests there would be merit in doing so.

Further research might test voter responses in the context of high-information experiments and tasks (Andersen and Ditonto 2018); and in less polarized political settings to assess whether moral licensing is less likely to arise when voters feel less strongly about the partisan choice. Future research could also offer a more detailed analysis of moral licensers, exploring the connection between moral licensing behaviour and other social attitudes such as hostile and benevolent sexism; measures of social conservatism; and wider system-justifying ideologies.

Conclusions

We investigated how voters reflect on a co-partisan candidate after being informed of sexual misconduct allegations against him. We identify two prominent strategies among voters who remain favourable towards the candidate: (1) moral licence is extended to the candidate; and (2) participants disbelieve and reject the allegations, pointing to reasons why they may be false or misleading. With the moral licensing strategy, our respondents refer back to the candidate’s background, professional track record and past political achievements that were outlined in the first vignette. These ‘positives’ appear to serve as credits to excuse the ‘negatives’ that emerge in the second vignette. The second important strategy is to simply reject the allegations as untrue, with scepticism justified by the timing of the allegations amongst other reasons. We examined whether voters who opt for moral licensing – the dominant of the two strategies identified – possess particular traits or characteristics. Within the limitations of our small sample, we suggest that moral licensers are less likely to report trust in people in general, and may report stronger partisan identity with the Democrat party.

Our qualitative analysis provides observable evidence to better understand the potential strategies employed that allow voters to justify looking past serious allegations of misconduct by co-partisan political candidates. We contribute to prior scholarship by offering in-depth thematic analysis of voters’ reasoning in their own words, as a complement to the largely quantitative survey experimental evidence on political scandal and misconduct. We also contribute to the moral licensing literature with a novel application of moral licence being extended to others, in the context of an important real world political phenomenon of sexual misconduct.

We identify a number of avenues for further research on how voters respond to political misconduct. Much of the work on political scandal relies on respondents based in the US, and there is a need to expand research to a broader cross-country perspective. Do responses to serious political scandals and/or sexual misconduct vary across different political and cultural contexts, and what role does moral licensing or disbelief play in different contexts? Is it possible to define more clearly the voters who turn to moral licensing or disbelief in political decisions? Are some candidates more likely to be given moral licence than others, and if so based on what factors? Finally, might there be spillovers from extending moral licence to a political leader on the voter’s own behaviour in the same or some other domains? In a context of increasing concern about democratic backsliding, we suggest that moral licence strategies might be investigated as a possible explanation for how voters look past violations of democratic norms.

Ethical approval

Approval was obtained from the ethics committee of Brunel University London (23716-MHR-Aug/2020-26640-1). The procedures used in this study adhere to the tenets of the Declaration of Helsinki.

Informed consent

Informed consent to participate in the study was obtained from all participants, including consent to publish non-identifiable data.

Acknowledgements

We thank Brunel University London and Royal Holloway University for research funding, and Luke Coughlan for superb research assistance. We greatly appreciate comments and feedback from panel participants at the Political Studies Association’s annual conference 2023 including Vibhor Mathur and David Redlawsk; participants at the Elections, Public Opinion and Parties annual conference 2023; the Loughborough Political Communications workshop 2023, with particular thanks to Will Allen for his insightful suggestions; and to Justin Fisher, Martin Hansen, Jeff Karp and Katja Sarmiento-Mirwaldt at Brunel’s Open Data Analysis Research Group. We would also like to thank three anonymous reviewers and the journal editors for their excellent suggestions and feedback.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets generated during and/or analysed during the current study are available from the corresponding author upon reasonable request.

Additional information

Funding

This work was supported by Brunel University London [grant number RDF Award LBG178].

Notes

1 Chris McGreal and Martin Pengelly (2023).

2 See polling by Fivethirtyeight through 2024 and spring 2024: https://projects.fivethirtyeight.com/polls/president-primary-r/2024/national/ and https://projects.fivethirtyeight.com/polls/favorability/donald-trump/.

3 The study was approved by a research ethics committee at Brunel University London (23716-MHR-Aug/2020-26640-1). Part of the study was a pre-registered online survey experiment which is published separately, see Savani and Collignon (2023).

4 This was a data-driven exercise rather than a theory-driven one (Hsieh and Shannon 2005), in line with this being an inductive rather than a deductive investigation; accordingly this analysis was not set out in a pre-analysis plan. The qualitative analysis is not corroborating a statistical hypothesis test, but focusing on the discovery of new or unexpected insights.

5 A similar approach is taken by Mathur (2023) when assessing voter reasoning towards a news vignette with Indian voters, with voters demonstrating more than one approach to resolving cognitive dissonance.

Appendix

Table A1. Descriptive statistics for survey respondents.

Variable	All survey respondents (n = 704)	Respondents aware of allegations and positively inclined towards candidate (n = 159)	Respondents aware of allegations and negatively inclined towards candidate (n = 210)
Age (modal)	30s	30s	30s
Female (%)	41	40	40
Married (%)	61	57	57
Education (modal)	UG degree	UG degree	UG degree
Holds politically liberal views (%)	53	47	60
Political knowledge (%)	78	76	80
Trusts in people in general (%)	54	56	43
Trusts in media (%)	70	71	76
Strong or very strong Republican partisan identity (%)	32	40	24
Strong or very strong Democrat partisan identity (%)	40	44	44
Self transcendence values score (0-5)	3.61	3.72	3.91
Self enhancement value score (0-5)	3.56	3.91	3.42

Notes: no quotas were applied to the sample.

Table A2. Thematic analysis of qualitative responses from voters who do not support the candidate after learning about the sexual harassment allegations (n = 210).

	Theme	Excerpts and examples of coded references
1	Sexual harassment allegations matter in the candidate evaluation (73%)	‘The former sexual activity' (id 416) ‘What he was accused of' (id 190)
2	Importance of character, morals, ethics and integrity of candidate (25%)	‘The lack of ethical behavior in harrassing others' (id 505) ‘Even if he has a good political record, it is most important to me how he treats and respects people personally. I think that shows true character' (id 89)
3	Trust in allegations and negative inference from lawsuit (27%)	‘Where there is smoke there is fire. The accusations raised flags (id 337) ‘People who settle lawsuits tend to be guilty' (id 295)
4	Candidate does not meet standards for public office (9%)	‘I believe that political leaders need to be an example in terms of moral and ethical behavior. It doesn't matter what party they are in.' (id 24)
5	Want to know more, finely balanced decision but not enough information to vote in favour (11%)	‘I would need to look into the case more and get details. I would also see who the opposing candidate is.' (id 313)

Notes: The proportion of text responses covered by this theme relative to the dataset is given in parentheses. Data analysed using NVivo 12.

Within theme 3, 47 respondents explicitly mention the lawsuit as a basis for their unwillingness to vote for the candidate (22.5% of this sub-sample of 209 respondents). These themes, particularly theme 1, are also discussed in Savani and Collignon (2023).