Harassment and Intimidation of Parliamentary Candidates in the United Kingdom

The use of political violence to attain political goals has long been a source of concern. Once thought to be exclusive to countries with high levels of general violence, recent evidence suggests that harassment and intimidation of political elites in the UK is more widespread than previously thought. Using data from the 2017 general election candidate survey, we ﬁ nd that four in every ten candidates experienced at least one type of harassment. Evidence suggests that women and young candidates are more likely to suffer from harassment and intimidation. We conclude by formulating an agenda for future research, focussing, in particular, on the perception of harassment and the effect of harassment on political careers.


Introduction
THE PHENOMENON of the harassment and intimidation of election candidates provides a challenge for political scientists and practitioners. It is a requirement of any liberal democracy that any individual legally able to stand for office will have an equal chance of being elected in free and fair elections, where their personal safety is guaranteed. However, this is often not the case. For example, in the UK, the murder of Labour MP Jo Cox in 2016 while campaigning to remain in the European Union, sparked concern about the personal safety of politicians. Later, in 2017, Sheryll Murray, MP for South East Cornwall, was the first member of the Conservative Party to speak out about her treatment during the campaign, saying it involved someone urinating at her office door and death threats on social media. She was not the only one experiencing aggressive behaviour; the online 'trolling' of prominent politicians, such as Diane Abbott, attracted major attention during the election campaign. These experiences of harassment and intimidation are different in their form of delivery and content, but it is undoubtable that they greatly affect candidates physically and emotionally. This is evident in discussions after the new Parliament was constituted, when the issue of harassment of candidates and MPs was raised by members from all sides of the Commons, showing that the intimidation of political actors in the UK was a source of concern across parties. But despite the political pressure and media attention to the problem, the situation did not improve in the following years. When a new general election was called in 2019, a string of politicians stepped down, among them Heidi Allen, MP for South Cambridgeshire, and Nicky Morgan, MP for Loughborough, citing abuse, harassment and intimidation as their main reasons.
This article contributes to our systematic understanding of aggression towards political elites, in particular towards candidates standing for office in the UK general election of 2017. Using original data, it shows how widespread harassment among candidates is and who is more likely to suffer from it.
There are two main strands of the literature that have sought to address the issue of aggression towards political figures. First, physical manifestations of harassment, such as stalking, have mainly been covered from a mental health and security perspective. As public figures, MPs are prone to attract the attention of people with mental problems 1 who become 'fixated' on prominent individuals. Following on from previous research on stalking and other forms of harassment of people with a high public profile, a major study of harassment and stalking of MPs was undertaken in the UK in 2010, and similar studies have been conducted in Australia, Canada, New Zealand, and Norway. 1 The second strand of the literature has focussed on the online harassment of politicians. As part of a large literature on the rise of social media and its effect on the nature of political discourse, a number of studies have analysed tweets and retweets directed at politicians in the UK and their 'incivil' and abusive content. 2 Other approaches include surveys of MPs regarding their experiences of online abuse. 3 Both strands of research conclude that the prominence and high visibility of politicians is one important predictive factor for all types of harassment-like that suffered by celebrities, film stars, and artists-and that both stalking and trolling are associated with psychopathological conditions of the perpetrator. We argue that there are still a number of important questions that deserve detailed examination from a political science perspective.
Firstly, we should not disregard the experiences of individuals who participate in politics but have not yet been able to build a prominent profile or win a seat in Parliament, since their experiences may play a key role in shaping their political ambitions or deter them altogether from future participation in politics. Thus, instead of focussing exclusively on sitting MPs, this article expands our understanding of harassment towards candidates in general. Second, a very important question concerns the effect of harassment on different types of candidates. Much of the public focus has been on female politicians, treating this as an issue of public life and institutionalised sexism in the workplace. 4 However, our understanding of this phenomenon is mainly based on anecdotal evidence and testimonials, lacking the systematic analysis that would allow comparisons. Are women more likely to be the subject of harassment than men? Do they suffer from distinctive forms of aggression? Are other groups also vulnerable to harassment? Third, some commentators see Britain following the USA and entering an 'age of incivility'. 5 The increasing role of online communication offering the choice for the user to hide their real identity could be seen as a contributing factor to a more 'incivil' type of discourse, and thus an increase in online harassment might be expected independent of any political polarisation. But what about other, offline forms of harassment? Are candidates also suffering from physical abuse? A final question relates to the effect of harassment on feelings of security among candidates. In this time of contentious politics, it is easy to dismiss claims of harassment as 'part of the job' or to believe that this issue is just used by some politicians to gain popularity. Are candidates really afraid for their personal security? Analysing variations in emotional responses is key if we want to identify the demographics of candidates most likely to alienate themselves from participating in politics.
The main aim of this article is to show that intimidation of political elites is indeed a problem that threatens the quality of democracy in the UK, and to identify those candidates that are more likely to suffer aggression. In order to fulfil our aim, we report results from an original large ESRCfunded survey that explores experiences of harassment among candidates standing in the 2017 general election in the UK. 6 The results show that: a) harassment and intimidation of parliamentary candidates are widespread across the UK; b) female, young and leading candidates are being targeted; c) harassment has emotional consequences for candidates; and d) there are differences in the frequency of online and offline harassment, but that these differences are more related to the content of the threat than the target of harassment. The rest of this article is organised as follows: first, by presenting the data obtained by the Representative Audit of Britain Survey and the methodology used for data collection; then by presenting the results organised around the research questions presented above; finally, the conclusion outlines avenues for future research.

Data and methodology
The analysis is based on an individual-level survey data collected as part of the Representative Audit of Britain Survey (RAB) between June 2017 and May 2018. We obtained 1,495 responses, corresponding to a 53 per cent overall response rate. The response rate by sex was 57 per cent females and 51 per cent males.
There is, as expected, variation in response rate by party. This variation can be seen in Table 1 below. We employed different techniques, made a great effort to ensure a good response rate and encouraged candidates from all parties to respond. However, we still found that party membership is a significant variable to explain non-response bias and harassment, as can be appreciated in Table 2, and we therefore, weighted results by party.

Defining and measuring harassment
Identifying or defining the nature of harassment and intimidation presents empirical and theoretical difficulties. Political scientists tend to define violence narrowly as an act of force. Sociologists, in contrast, classify violence more comprehensively to include actions that are violent but normalised by society. 7 This means that asking candidates about harassment using a stringent definition will prompt responses that refer only to widely recognised forms of abuse, leaving out other more subtle experiences most likely suffered by women. In consequence, we deliberately avoided providing a definition of harassment or narrowed its scope in the questionnaire. Instead, we asked candidates openly if they personally experienced any form of inappropriate behaviour, harassment, or threats to their security in their position as parliamentary candidates during the campaign.
Some criticism of this approach stems from the fact that we are looking at perceived acts of intimidation and that there is likely to be variation between what individuals consider harassment to be. However, we stand by our definition because it appeals directly to the sense of security of the individual rather than abiding by some normative definition of what does or does not constitute a threat by an external observer. Moreover, we argue that candidates, as any other human being, react to perceived threats whether they are real or not. Approximately four in every ten candidates suffered intimidation during the campaign, with 38 per cent of candidates answering positively to the question. A simple cross-tabulation with a Pearson's v2 test shows that women are particularly targeted, with 45 per cent of female candidates suffering harassment and intimidation, compared with 35 per cent of males (p<0.05).

Findings
Who is being targeted? Are women more likely to suffer from harassment and intimidation than men? What about other factors that may put other candidates at risk? Results from a logistic   We added a measure for how competitive the race was by including a dummy variable coded one for candidates who finished first or second in the race and zero otherwise and the margin of majority in the constituency. Of course, we understand that both measures are post-election measures that only reflect how the campaign ended, but not its development. However, we believe that the competitiveness of the campaign and the visibility of the candidate will be reflected. The larger the margin of victory, the less competitive the campaign.
Results indicate that candidates standing for the Greens, Liberal Democrats, Labour and Plaid Cymru are significantly less likely to be harassed than candidates standing for the Conservative Party (reference category). However, as can be appreciated in Tables 1  and 3 below, there is a negative relationship between the response rate and the per cent reporting harassment, which could arise if candidates from the Conservatives, UK Independence Party (UKIP) and Scottish National Party (SNP) who were not experiencing harassment chose not to respond to the survey. Thus, it is possible that these differences in party response rates could affect this finding.
The coefficients show that female, young and leading candidates are significantly more likely to suffer harassment than their counterparts. Marginal probabilities indicate that women are 9 percentage points more likely than men to suffer harassment. Younger candidates are more likely to become victims of intimidatory behaviour and this likelihood decreases with age. For example, a twenty year-old candidate is 18 percentage points more likely to be harassed than a candidate who is fifty years old. Leading candidates are 9 percentage points more likely to be harassed than non-leading candidates.
Is online harassment more or less frequent than physical abuse? Who suffers from it? We asked candidates specifically what type of abuse they suffered and how often it occurred. It can be seen in Figure 1 below that the most frequent abuse occurs on social media; 29 per cent of the total number of candidates confirmed they received improper communications on Twitter, Facebook, and so on, at least once during the campaign. Women are targeted (34 per cent) more often than men (26 per cent). The second most frequent form of harassment is intimidation via email, with 23 per cent of candidates receiving inappropriate messages at least once during the campaign. Again, this percentage is significantly higher for women (28 per cent) than men (20 per cent). Women suffer significantly more threats of harm (5 per cent reported having been threatened three or more times during the campaign compared with 3 per cent of men) and from individuals loitering around their home or other places they frequent (4 per cent of women experienced these three or more times compared with 2 per cent of men). The least frequent form of abuse-sexual harassmentalso presents differences by sex. In general, 1 per cent of the total number of candidates had been victims at least once, but this percentage is entirely formed by women (3 per cent of female candidates), as men did not report any instance of abuse in this category. We do not find evidence of targeting by sex in the remaining categories. In order to determine if certain candidates are more likely to be abused in person or online, we proceeded to fit a series of tobit regressions using as the dependent variable for the first four forms of intimidation identified as more frequent. All dependent variables are measured with six distinctive categories (No, Once, Twice, 3-5 times, 6-9 times and more than 10 times). The models include the same independent variables described before. Table 4 below shows the results of this series of models (models 3 to 6). From the four categories of harassment used, women reported a higher frequency of abuse on social media and over email. But when controlling for other characteristics, women do not suffer more unwanted approaches and threats than men. Incumbent MPs did not receive significantly more threats than other candidates on social media, but did receive more emails and were more frequently approached or threatened during campaigns than others. With the exception of threatening emails, leading candidates received significantly more harassment in all forms than non-leading candidates.
We found that young candidates were more frequently harassed and intimidated in any form than older ones as the coefficient of age is significant and negative in all cases. BME candidates received fewer threats on social media and over email, but the significance is marginal, reflecting the low number of BME candidates who responded to the survey.
The party coefficients suggest that candidates standing for the Conservative Party (baseline category) receive significantly more abuse in any form than candidates standing for other parties, with the exception to UKIP candidates, who are more often approached and threatened than Conservatives. However this finding should be read in the light of potential biases introduced by party differences in response rates, as noted above. Does harassment affect all candidates equally? Do they fear for their security? Many psychological scientists now assume that emotions are the dominant driver of decision making, because decisions serve as a way of avoiding negative feelings such as anger or fear. This suggests that harassment will have a stronger impact on the behaviour and political ambitions of candidates who are emotionally affected by it. To investigate this issue further, we asked the candidates to indicate to which intensity they felt fearful as a result of the harassment experienced, with 32 per cent of candidates affirming that they felt either moderately of very fearful as a result of their experiences of harassment and intimidation.
We investigated this issue further by fitting another tobit model where the dependent variable was feelings of fear. We asked the candidates if, as a result of their experiences of harassment and intimidation, they felt fearful, giving them five options: 0 = did not suffer harassment; 1 = suffered harassment but felt no fear at all; 2 = only a little; 3 = moderately; and 4 = very fearful. The list of coefficients is presented in Table 3, column 2 (M2) above. In this case, the coefficients indicate that female and incumbent MPs are more fearful than their counterparts. On the other hand, older candidates experience significantly less fear. With the exception of SNP and UKIP candidates, parties standing for all Candidate visibility is important for all types of harassment we analysed. The fact that incumbent MPs and leading candidates are at a significantly higher risk of harassment online and offline indicates that, while social media makes communication easier, this is not the main reason why abuse is on the rise. It is possible instead that, as some police reports suggest, a rise in the abuse and harassment of prominent politicians can be explained by increasingly polarised opinions on Brexit. In terms of gender, we found evidence that women are being targeted, as they suffer significantly more harassment, mainly on social media and via email. The fact that they express fear as a reaction to their experiences of intimidation indicates that we should look into the content of the threat as well as the form of delivery it takes.
Age comes across as an important predictor applying to all forms of harassment of election candidates. The reasons for this are not immediately clear and need further investigation. It is possible that younger candidates have a different understanding of what constitutes harassment, or that older candidates have normalised the aggression that is aimed at them. There is some indication that this may be the case, as age generally reduces feelings of fear.
Some of our results raise new questions. Variations in the type of threats suffered by candidates suggest a gender and generational element as well as differences in the mechanisms behind online and offline abuse. While the focus on online harassment has dominated the recent news and research agenda, research should also continue into offline forms of abuse which may have a more direct impact on candidates.
What clearly is a major issue for future research is the effect of harassment on candidates' political careers. Our results suggest that young female candidates who commit themselves to campaigning in a major way are the most affected by harassment and intimidation during election campaigning. There is plenty of evidence from studies of bullying and workplace harassment that this has a major emotional effect on the victims of such behaviour. This is a topic of intrinsic political importance if, as a result of harassment, people are deterred from becoming active in politics or are effectively pushed into abandoning their political career before it has really started.