key: cord-0779543-ixkx35ot
authors: Herrick, Rebekah; Thomas, Sue
title: Violence among State House Candidates during the COVID‐19 Pandemic
date: 2021-11-01
journal: Legislative Studies Quarterly
DOI: 10.1111/lsq.12361
sha: 030947e8970c8e8821252d39749a257c873492c0
doc_id: 779543
cord_uid: ixkx35ot

Although there is an established comparative literature addressing electoral violence, and an emerging literature on violence against US officeholders, research has not examined levels of violence perpetrated by constituents against US candidates, nor how a crisis, such as the COVID‐19 pandemic, including the political polarization connected to it, may affect violence. This note presents findings from a survey of state house candidates from 18 US states in the 2020 electoral cycle. The findings focus on psychological violence and suggest that it is a common occurrence, and that the COVID‐19 pandemic and polarization may have affected it. Specifically, most candidates reported that: they believe that the pandemic led to violence; those in states with larger increases in COVID‐19 cases and those running in states with higher levels of COVID‐19 public political events reported more violence. In addition, candidates running in more polarized states experienced more violence.

i received quite a bit more harassment this time than when i ran for this office in 2018 but i am not sure why. i ran a more aggressive campaign this time, the pandemic has people on edge, and our president models the inappropriate behaviors i received. …. i had my campaign literature stolen and spread in supporters' yards. i received phone calls, anonymous snail mail, and hateful emails and messages. that didn't happen last time. But was COvid the cause? -survey respondent in this research note, we address the degree to which candidates for state house races in the united states have experienced violence from prospective constituents. Our focus is on how candidates perceive the association between COvid-19 and polarization related to it and the psychological violence experienced during campaigns. using a survey of candidates from 18 us states during the 2020 electoral cycle, this study is unique in its focus on violence against candidates perpetrated by constituents and its emphasis on psychological violence, defined as actions likely to cause psychological harm. this includes a broad range of experiences from social media and verbal abuse to threats against candidates and their families. Psychological violence is also more common than physical attacks (thomas et al. 2019) . understanding electoral violence, including psychological violence, is important as it can limit willingness to run for office (Krook 2020) , contribute to psychological stress , depress voters' level of political knowledge (söderström 2017), lead to worse politicians and corruption (dal Bó et al. 2006) , and undermine citizens' trust in the electoral process (Birch 2020, chap. 1) .

Our work is informed by three strands of political science literature. First, comparative studies that examine the antecedents and effects of electoral violence provided a general foundation on the subject including the connection between political polarization on political violence. this literature makes clear that state-sponsored or mass violence against the general population, candidates, journalists, and others varies in form, conditions, and contexts (Bardall 2011; Birch 2020; Hafner-Burton et al. 2014; Harish and toha 2019; LeBas 2006; Mochtak 2018) .

However, as our work focuses on individually initiated psychological violence against us candidates rather than statesponsored organizationally directed violence, the second foundation of our research comes from studies of us officeholders. the emerging literature on us electoral violence aimed at officeholders prior to COvid-19 finds that more than 80% of both us mayors and us state senators have faced psychological violence inflicted by constituents thomas et al. 2019). Further, Herrick, thomas, and Bartholomy (2021) report that political polarization is associated with increased violence against legislators by their colleagues.

another emerging strand of both comparative and us literature pertains to . studies report that polarization worsened in the united states during the pandemic (gadarian et al. 2021; gollwitzer et al. 2020; Warshaw et al. 2020) . For example, gollwitzer et al. (2020) found that partisan differences in attitudes concerning COvid-19 grew in 2020 as the pandemic worsened. Connecting political polarization, electoral violence, and the pandemic, Byman and Clarke (2020) find that polarization is associated with electoral violence, including against candidates. additionally, a recent branch of comparative research centers on the effects of COvid-19 on elections and violence -and suggests an association between COvid-19 lockdowns and civil protests and violence (gutiérrez-romero 2020; James 2021).

these strands of political science literature can be augmented from social science literature on general workplace violence. although being a candidate is different from typical workplaces, there are elements of this literature that can inform the prevalence and correlates of these experiences in public roles. What we find is that psychological violence is fairly common, and that public administrators and government employees face high rates of these behaviors (Baron and Neuman 1996; Fischer et al. 2016; Harrell 2013; schat et al. 2006) . Further, one of the most consistent factors associated with employee violence appears to be the amount of contact workers have with the public ( Baron and Neuman 1996; Fischer et al. 2016; Piquero et al. 2013) .

altogether, following the lessons of the comparative and us literature on political violence, polarization, and pandemic effects, we investigate the degree to which us candidates for state house races in the 2020 electoral cycle reported episodes of psychological violence and whether candidates perceived that COvid-19 is associated with increased psychological violence. examining the beliefs allows for a subjective as well as an objective measure about the effects of COvid-19. in specific, we explore the following relationships:

Candidates running in states with higher levels of COvid-19 infections in relationship to levels of psychological violence.

Candidates running in states with more COvid-19-related public political events, such as protests, levels of psychological violence.

Candidates running in more polarized states and levels of psychological violence.

in 2020, we conducted a survey of state house candidates in 18 states -alaska, arkansas, Colorado, delaware georgia, illinois, iowa, Kentucky, Maine, Minnesota, Montana, Nevada, New Mexico, Oklahoma, Oregon, Pennsylvania, rhode island, and Wisconsin. these states were selected for diversity among a host of variables including political culture (elazar 1966), region, 1. party control of the lower chamber, 2. legislative professionalism, 3. and legislative accountability. 4. surveys were sent in November and december to 2020 general-election candidates 5. in these states who ran in races that featured a democrat and republican candidate. 6. the survey was mixed mode (mail and internet). Candidates received up to five contacts: (1) an email with a Qualtrics link; (2) an email reminder with a link; (3) a mailed survey; (4) a postcard reminder; and (5) a final email with a link. the response rate was 28% (617/2,208), which is higher than most recent surveys of elites in the united states (Hanania 2017; Purtle et al. 2019; thomas et al. 2019) . to detect biases in responses, we checked differences in response rates by gender, party, incumbency, professionalism of the legislature, political culture, whether the state was a battleground state, 7. and average district population size. although there were a few statistically significant biases, substantive significance was small -except regarding incumbency. 8. Whereas 34% of nonincumbents returned surveys, only 20% of incumbents returned them.

We use two dependent variables: the frequency of psychological violence and candidates' beliefs about the pandemic's effect on violence.

Psychological Violence 9. . to measure the levels of psychological violence experienced by candidates, we created on index based on the frequency with which candidates reported the following:

• Harassment • awareness of content in social media about the candidate that was untrue and offensive • awareness of content in the traditional media about the candidate that was untrue and offensive • awareness of content from a public event about the candidate that was untrue and offensive • threat(s) of death, rape, beating, or similar act to the candidate • threat(s) of death, rape, beating, or similar act to the candidate's family For each of these items, candidates were asked to indicate the frequency of these experiences: 1 (never), 2 (less than monthly), 3 (once or twice a month), 4 (three or four times a month), or 5 (more than four times a month). the resultant index ranged between 6 and 25 (α = 0.75).

Belief that COVID-19 Increased Violence. to measure our second dependent variable, we used answers to this question: "How do you think that COvid-19 pandemic affected levels of aggression?" the options ranged from 1 (greatly decreased levels) to 5 (greatly increased levels). the 81 respondents who said they were unsure were treated as missing. this variable is used for two reasons. the first is to serve as a proxy for a baseline measure of change as experienced candidates may have a good idea of changes in their district. the second is that understanding candidates' beliefs about how the pandemic affected levels of aggression is important as perceptions guide behaviors.

to measure the effects of COvid-19 on these behaviors, we use two state-level variables. the first variable, increases in COvid-19 cases, is changes in the number of daily COvid-19 cases between October 1 and November 1, 2020, in each state. 10. this variable measures per capita cases and ranges from 9 to 437.97. the second independent variable, COvid-19 public political events, is the number of COvid-19 related events reported by the armed Conflict Location and event data (aCLed) project. 11. this variable is also measured per capita and ranges from 0 to 4.29.

the independent variables measure different ways the pandemic may have affected levels of violence reported by candidates and they are not collinear. the correlation between the number of cases and the number of COvid-19 public political events is 0.03 (p = 0.48).

Our third independent variable, polarization, comes from shor's (2020) measure, which is based on differences in the ideological stances between the parties in the state legislatures. the measure was available for 2018 for every state but iowa. Hence, we relied on iowa's score in 2017. 12. in this dataset, it ranges from 0.66 to 2.91.

Following research examining violence against us mayors and state senators (Herrick and thomas 2021; thomas et al. 2019), we include control variables to account for how violence is related to individual traits (age, sex, race, and party), 13. district traits (population, density, socioeconomic status, and age), 14. state traits, (traditional political culture 15. ), and legislative professionalism. 16. We also control for electoral factors that may affect the levels of violence against candidates (competitiveness of the election and incumbency) since research on electoral violence finds that competition affects its levels (Birch et al. 2020, 7) . 17. Finally, there were other significant political influences in 2020 that may be associated with levels of violence including the trump reelection campaign, the Black Lives Matter movement, and related police and race-based public political events. to account for these influences, we used the aCLed data to develop measures of the number of non-COvid-19 public political events. this variable was transformed to the number of events per million residents. 18.

We use mixed effects linear regression to analyze psychological violence 19. and mixed-effects-ordered logistic to analyze beliefs about the effects of COvid-19 on violence. For both, we weighted the data by incumbency. violence. Overall, 54% reported that COvid-19 increased violence, and 46% felt it had no effect, decreased it, or were unsure of the effects of the pandemic on violence. although candidates believed that COvid-19 increased violence, there was little indication that state-level COvid-19 cases, COvid-19 public political events, or polarization were related to this belief (see column 1, table 2). although the number of COvid-19 public political events and general levels of polarization were correctly signed, neither they nor the COvid-19 cases variable were statistically significant. thus, state-level COvid-19 conditions were not related to candidates' beliefs about levels of violence. it may be however, that either district-level or national-level indicators of COvid-19 are more likely to affect beliefs about the effects of COvid-19 on violence against candidates for state house seats.

although not the main topic of this note, we also report on the extent to which each control variable in our model affected candidates' beliefs about the effects of COvid-19 on experiences of violence. Our data suggest that the following types of candidates were more likely than their counterparts to believe that COvid-19 affected levels of violence: those running in less populous districts, more densely populated districts, districts with higher socioeconomic status levels, those running for seats in more professionalized legislatures, incumbents, and those running in states with fewer non-COvid-19 political protest events.

Psychological violence was common among our respondents: 84.9% of candidates reported at least one episode. as in prior research thomas 2021; thomas et al. 2019) , the primary conduits of violence were social media. social media abuse averaged 2.86 on the 5-point scale, and harassment averaged 2.47. in addition, two-thirds of respondents reported harassment and three-quarters reported social media violence.

table 2 shows that levels of psychological violence are associated with the rise in COvid-19 cases and number of COvid-19 public political events (see Column 2). For example, the difference between the average candidates' reported experience of psychological violence between those who ran in states with the largest and smallest increase in COvid-19 cases is 1.71 (SD = 4.37). similarly, the difference between the average candidates' reported experience of psychological violence between those who ran in states with the most and fewest number of COvid-19 public political events is 1.76.

table 2 also shows that polarization is related to violence against candidates. the difference between the average candidates' reported experience of psychological violence between those who ran in states with the highest and lowest levels of polarization is 3.19. although this is not a measure of COvid-19-related polarization, it indicates that violence is related to polarization. Note: Numbers in parentheses are standard errors. ***p < 0.01; **p < 0.05; *p < 0.10 (using a two-tailed test).

Factors other than the pandemic were related to candidate's experiences of psychological violence. in terms of personal traits, only age, incumbency, and party were statistically significantly related to violence. young candidates, incumbents, and republicans reported more of these experiences. in addition, candidates running for professional legislatures and in states with traditional political cultures reported more violence than their counterparts. Only one district trait was related to levels of violence: median age. the older the median age of a district, the more likely its candidates experienced violence. Finally, those running in competitive races reported more than others.

in all, we found that: (1) candidates believe that the pandemic led to violence against them, (2) candidates running in states with larger increases in COvid-19 cases reported more psychological violence, (3) more COvid-19 public political events were associated with more psychological violence, and (4) candidates in more polarized states reported more psychological violence. On the other hand, candidates' beliefs about the effects of COvid-19 were not associated with COvid-19 rates, COvid-19 public political events, or polarization.

to summarize, we found that candidates believe that the pandemic led to violence against them: more than 50% of candidates perceived that COvid-19 increased these episodes. Further, almost 85% of us state house candidates in 2020 reported experiences of psychological violence. Our results are consistent with the electoral-violence research literature that context matters. the presence of the pandemic was associated with candidate reports of increased levels of violence: those in states with larger increases in COvid-19 cases and a larger number of pandemic-related public political events reported higher levels of violence. at least part of the reason that the presence of COvid-19 was associated with higher levels of psychological violence may be the polarization of politics in the united states, including or especially, polarization about pandemic policy. this conclusion is indirectly supported by our finding that state legislative polarization is related to high levels of psychological violence.

the results of the effects of the control variables in our models may also be informative for future research. First, although district-level variables were related to perceptions about the effects Violence among State House Candidates during the COVID-19 of COvid-19 on psychological violence, apart from median age, they were not related to levels of psychological violence. second, power may affect levels of psychological violence experienced by candidates. that is, those running for professionalized legislatures and as incumbents reported more violence than their counterparts. third, although republicans reported more psychological violence than democrats, there were not significant party differences in beliefs about the effects of COvid-19 on violence. similarly, those running in competitive races experienced more psychological violence but did not differ from others in their belief about the effects of COvid-19 on violence. Finally, in contrast to the literature on violence against us officeholders, women candidates did not report significantly higher levels of psychological violence than men. two reasons may be that COvid-19 equalized the gender playing field in 2020 in the united states, or that gender affects violence more for sitting officeholders than candidates.

in all, our study extends the research on polarization on electoral violence by adding several elements: (1) evidence of the effects of context (the global pandemic) on electoral violence; (2) finding that polarization is associated with psychological violence against candidates as well as officeholders and mass publics; (3) evidence of polarization on electoral violence in the us case in addition to understandings of it in developing democracies; and (4) evidence of the violence perpetuated by constituents as well as that perpetuated by political leaders. Further, the findings suggest that although polarization is associated with higher levels of psychological violence, it is not associated with the perception that COvid-19 increased violence. it may be that baseline levels of psychological violence were high enough in these states that COvid-19's effects were not strongly felt.

although our survey is pathbreaking and informative, it is not without limitations. First, it is possible that those experiencing psychological violence were more likely to complete the survey than others. On the other hand, it is also possible that those experiencing violence were less likely to respond because they did not want to relive difficult events. However, even if the percentage is not exact, it is probably safe to say psychological violence is common on the campaign trail. second, response bias is a common problem in survey research. as an example, what one candidate might consider threatening or offensive may be seen as minor to another. although we tried to minimize this problem by asking about specific types of actions, variation in how candidates interpreted events is likely. 20. relatedly, individuals' willingness to report being a victim of violence also varies. to minimize this possibility, we were careful to avoid using charged terms like "victim of violence" in our survey. Nevertheless, these issues may remain. third, although surveys are not perfectly accurate reports of episodes of psychological violence, they produce superior data compared to official complaints or newspaper accounts. there is likely even greater variation in the likelihood of formal reporting of these encounters and differences in whether formal reporting is covered by various media (Håkansson 2021) . Fourth, there are other ways of measuring the central concept of our work -psychological violence against political candidates. For comparability across existing measures of this concept among political elites inter-Parliamentary union 2016; thomas et al.-2019) , we chose the measures reported here. Future researchers may want to expand the measures and supplement them with qualitative analysis. Finally, although our control variables cannot guarantee more than a correlation between COvid-19 and violence, our models control for the main factors found in the literature that affect violence, such as institutional strength (measured by professionalism -see Birch 2020). 21. it is possible that COvid-19 indicators tap other state characteristics related to violence. yet, since our two measures of COvid-19 are not strongly related to each other, and we control for numerous state-level conditions, this may be unlikely. in all, a call for more research to better explore the findings reported here is even more critical than usual. as this is the first study that explores violence against us candidates at the hands of constituents, and it was conducted during a global pandemic, much more work is needed. trends over time, levels of office, races during midterm versus presidential election years, experiences of psychological versus physical violence, state sponsored or citizen-perpetuated violence, and much more must wait for future studies. 22. 

9. the survey also included questions to measure levels of physical violence, but initial analysis suggested that these items did not fit together as a single construct. although we considered creating one variable that include psychological and physical violence, we opted to report psychological violence here. the reason is that previous research on officeholder found that these two types of violence are different in many respects (see, for example, thomas et al. 2019). the alpha of our selected measure was slightly larger without the physical violence variable in the index (0.736 to 0.753).

10. the correlation between this variable and the number of COvid-19 cases in October 1 is 0.66; for November, it is 0.93. these data come from the New York Times; see https://www.google.com/searc h?clien t=firef ox-b-1-d&q=covid +cases +by+state +octob er+2020.

11. these data come from the armed Conflict Location and event data project (aCLed) found at https://acled data.com/#/dashb oard). aCLed includes "political violence and protest" events. according to aCLed: "this data file contains disorder events that are directly related to the coronavirus pandemic. it includes events such as violence targeting healthcare workers responding to the coronavirus, mob attacks on individuals due to fears of their alleged links to the coronavirus (e.g., Muslims in india; foreigners in africa; etc.), demonstrations against governance decisions made in response to the coronavirus, and more" (https://acled data.com/curat ed-data-files/). We did not limit our analysis with any of their filters so it would include battles, violence against civilians, explosions/ remote violence, riots, protests and strategic developments, but most of the events used here would be classified as riots, protests or strategic developments.

12. this measure predates COvid-19 but can be used to indicate that polarization is related to electoral violence.

13. respondent's sex: open-ended question: "What is your gender?" race: open-ended ("What is your race?". Please list all races that apply and indicate if you are Hispanic"). Forty-four respondents left this blank or said something like "human race." We were able to determine race based on other publicly available sources for 13 candidates. We treated the remaining 31 respondents as white. as 90% of the respondents identified as white, the variable "white" is coded 1 if candidates identified as white only, and 0 otherwise. to measure age, the survey asked: "in what year were you born?" converted to age. there were 20 missing cases. For these, we used mean substitution.

14. the measures are median age, and population in thousands, people per square mile, district ses (the average of combining z-scores of percent college educated and median household income) and come from https://censu srepo rter. org/.

15. We included the traditional political culture measure only due to multicollinearity with moralistic and individualist cultures -and some state-level measures, particularly polarization and professionalism.

16. Professionalism: whether the legislature is full-time or part-time using NCsL's measure.

17. Competitiveness of the election is the percent vote and subtracted from 50, and then we converted that score to absolute values. smaller scores indicate more competitive elections. Percent of vote from state websites and Ballotpedia. incumbency from Ballotpedia.

18. We separated these into different categories of events such as BLM, Pro-trump, Pro-Biden. the total events variable had the strongest relationship. We also considered controlling for term limits since it is a common control in the state legislative research, but it was not significantly related and was dropped from the model due to overfitting.

19. We treat the psychological variable as linear because of the large number (20) of possible scores. 20. in the realm of gender politics research, some scholars do not classify this as problematic as it is individuals' interpretation of events that give them meaning (Krook 2020) .

21. two notable omissions in our set of control variables are ethnic diversity and levels of corruption (Birch, 2020; Kuhn 2015) . We tried to measure ethnic diversity with racial makeup of the district and levels of corruption, but they were not significantly related to violence.

22. data are available from authors upon request.

Legislative accountability: 21 indicators that tap citizen access to information and ability to participate: the degree to which legislators are prohibited from profiting in office, including nepotism; disclosure of conflicts of interest and limitations on those interest; disclosure of contact with lobbyists; revolving door limits; and judiciary review of related laws. see https://publi cinte grity.org/ state -polit ics/state -integ rity-inves tigat ion/how-does-your-state -rank-for

We only sent surveys to democrats and republicans. initially, we gathered names for candidates from Ballotpedia and cross-referenced them with state websites. 7. the battleground states were iowa

53); moralistic political culture r = 0.05 (p = 0.01)

Workplace violence and Workplace aggression: evidence on their relative Frequency and Potential Causes

electoral violence: an introduction

Electoral Violence, Corruption and Political Order

Why the risk of election violence is High

Plata o Plomo?': Bribe and Punishment in a theory of Political influence

American Federalism: A View from the States

Workplace aggression toward Local government employees: target Characteristics

Partisanship, Health Behavior, and Policy attitudes in the early stages of COvid-19 Pandemic

Partisan differences in Physical distancing are Linked to Health Outcomes during the COvid-19 Pandemic

Conflict in africa during COvid-19: social distancing, Food vulnerability and Welfare response

When do governments resort to election violence?

do Women Pay a Higher Price for Power? gender Bias in Political violence in sweden

the Personalities of Politicians: a Big Five survey of american Legislators

a New typology of electoral violence: insights from indonesia

Special Report BJS Workplace Violence against Government Employees

is it safe to Keep this Job? the Costs of violence on the Psychological Health and Careers of us Mayors

gender, Power, and Colleague aggression in us state senates

Physical violence and Psychological abuse against Female and Male Mayors in the united states

Not Just sticks and stones: Psychological abuse and Physical violence among u.s

New development: running elections during a Pandemic

do Contentious elections trigger violence?

Violence against Women in Politics

Polarization as Craft: Party Formation and state violence in Zimbabwe

Fighting and voting: Mapping electoral violence in the region of Post-Communist europe

state Legislators' Opinions about adverse Childhood experiences as risk Factors for adult Behavioral Health Conditions

aggregate state Legislator shor-McCarty ideology data

Fear of electoral violence and its impact on Political Knowledge in sub-saharan africa

Prevalence of Workplace aggression in the u.s. Workforce: Findings from a National study

Not for the Faint of Heart: assessing Physical violence and Psychological abuse against us Mayors

Fatalities from COvid-19 are reducing americans' support for republicans at every Level of Federal Office