In 2018, state senator Kiah Morris of Vermont resigned her seat amid in-person and social media threats and attacks. In 2016, she and her family began experiencing harassment, stalking at a polling place, and violations of her home and property. Staff members were also harassed. In the two years prior to her resignation, police responded to at least 16 calls from the senator and her family members (Aloe and McCullum Reference Aloe and McCullum2019). Unfortunately, Morris is not alone is her experiences of psychological abuse and physical violence from constituents.
A fast-growing theoretical and empirical literature illuminates the extent and consequences of psychological abuse and physical violence faced by women citizens, voters, and activists worldwide. Recent well-publicized cases of violence against women officeholders highlight this additional barrier to gender representation (see, e.g., Bardall, Bjarnegård, and Piscopo Reference Bardall, Bjarnegård and Piscopo2019; Håkansson, Reference Håkanssonforthcoming; Krook Reference Krook2017; Krook and Restrepo Sanín Reference Krook and Sanín2020). Knowing whether women in office face more or different types of abuse and violence than their men counterparts is essential to fully understanding the obstacles to political equality.
At present, though, scholarly research on these questions is scant, particularly in the United States. The only other study of the subject of which we are aware pertains to mayors (see Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). That study found that women were more likely than men to experience abuse and violence. To augment this nascent literature with evidence from the state government level, we report results from an original survey of all women and men state senators in the United States about their experiences of psychological abuse, physical violence, and sexualized abuse and violence from the public. Our aim is to examine the frequency, types, and correlates of experiences of abuse and violence among senators, with a specific focus on gender differences.
This body of work is important for several reasons. The consequences of officeholder vulnerability include limitations on policy makers’ ability to serve fully, make their voices heard, and accomplish their goals. Effects are also possible on commitments to continue public service. Experiencing abuse and violence increases the costs of service, which could depress political ambition. Similarly, learning about the costs of being a political representative may deter others from seeking office. Most central to this study, in situations in which women are subjected to greater levels or types of abuse and violence, the cost to serve may be higher than for men. Hence, a decrease is possible in the quality and diversity, especially the gender diversity, of those who lead us. Learning the contours of the problem is overdue.
Overall, our results indicate that more than 80% of state senators reported having faced abuse and violence and that women senators experienced more physical violence than men. We also found that different factors contributed to abuse and violence among women and men senators. Most notably, women with higher levels of power (party or committee leaders) were more likely than others to experience psychological abuse and sexualized abuse and violence; women in more professionalized legislatures reported more physical but less sexualized abuse and violence than their counterparts; and Democratic women senators experienced more sexualized abuse and violence than Republican women.
DEFINING ABUSE AND VIOLENCE
Our focus on psychological abuse and physical violence follows the comparative theoretical and empirical literature on obstacles to women's rights to full, free, and safe participation in political processes (Bardall Reference Bardall and Thompson2018; IPU 2016, 2018; Krook Reference Krook2017; Krook and Restrepo Sanín Reference Krook and Sanín2020). Both psychological abuse and physical violence involve what Bjarnegård (Reference Bjarnegård2018) refers to as violations of personal integrity. A constituent of each type of violation is sexualized abuse/violence, which is defined as unwanted physical behavior, such as sexual assault, rape, or exposing one's self without consent. Sexualized abuse includes statements, gestures, and images of a sexist nature, including sexualized threats (IPU 2016). Each of these three types of behaviors has been found to harm politicians’ psychological well-being and willingness to serve (see, e.g., Collignon and Rudig Reference Collignon and Rudig2020; Herrick and Franklin Reference Herrick and Franklin2019). Further, each type of abuse and violence may be explicitly gendered. Woman may experience these behaviors in different ways than men as specific acts may take gendered forms: female and male officeholders may process the treatment differently, and the effects of those experiences may affect their political careers in distinctive ways (Ballington Reference Ballington2018; Biroli Reference Biroli2018; Bjarnegård Reference Bjarnegård2018; Krook Reference Krook2017; Krook and Restrepo Sanín Reference Krook and Sanín2016; Kuperberg Reference Kuperberg2018; NDI 2016).Footnote 1
Based on the foregoing literature and following two surveys used by the Inter-Parliamentary Union, we define psychological abuse and physical violence in the following ways: Psychological abuse involves acts likely to harm the psychological well-being of individuals or their families by inducing fear or harm to their sense of self-worth or well-being. Physical violence involves activities that directly harm one's physical well-being or property. Specific operationalizations of these definitions are introduced later in our discussion of methodology.
LITERATURE REVIEW
Little scholarly research has been conducted on officeholders’ experiences of psychological abuse and physical violence overall and with regard to gender. We situate the few studies of officeholder abuse and violence in the broader literature on gendered workplace abuse and violence in the United States and in the comparative politics literature on the subject.
In the United States, a host of studies show that women in the general workforce face more of some types of abuse and violence than men (Berdahl Reference Berdahl2007; Holland and Cortina Reference Holland and Cortina2013, Reference Holland, Cortina, Connerley and Wu2016; Kabat-Farr and Cortina Reference Kabat-Farr and Cortina2014; McDonald Reference McDonald2012; Quick and McFadyen Reference Quick and Ann McFadyen2017). For example, 41% of women report having encountered workplace harassment in their lifetimes compared to 32% of men (Das Reference Das2009). Another example comes from a December 6, 2017, Quinnipiac University poll that reported that, among adults, 32% of respondents had been assaulted (17% of men and 47% women), and among those, 37% reported that the assault had happened at work.Footnote 2
Available research on officeholders tends to focus on women without comparisons to men—and few address the U.S. case. In fact, we are aware of only one U.S. study that compares the experiences of women to men (Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). In this research, the authors found that U.S. women mayors were more likely than men to experience psychological abuse and physical violence, as well as sexualized abuse and violence. The results also indicate that women who lived in traditional political cultures and were “strong mayors” (had veto and appointment powers) were more likely to experience psychological abuse than other women mayors.
Substantially more comparative research is available on this topic than in the United States. One group of studies focuses on women only and indicates that sexism, abuse, and violence against women parliamentarians is widespread (Bardall, Bjarnegård, and Piscopo Reference Bardall, Bjarnegård and Piscopo2019; IPU 2016, 2018; Krook Reference Krook2017; Krook and Restrepo Sanín Reference Krook and Sanín2020; Schneider and Carroll Reference Schneider and Carroll2020). Among the research studies that focus on gender comparisons (Bjelland and Bjørgo Reference Bjelland and Bjørgo2014; Bjørgo and Silkoset Reference Bjørgo and Silkoset2018; Every-Palmer, Barry-Walsh, and Pathé Reference Every-Palmer, Barry-Walsh and Pathé2015; James et al. Reference James, Sukhwal, Farnham, Evans, Barrie, Taylor and Wilson2016), authors report that women face more abuse and violence than men. Further, women report more negative experiences in each specific type, especially sexualized abuse and violence (Bardall Reference Bardall and Thompson2018; Bardall, Bjarnegård, and Piscopo Reference Bardall, Bjarnegård and Piscopo2019; Bjelland and Bjørgo Reference Bjelland and Bjørgo2014; Bjørgo and Silkoset Reference Bjørgo and Silkoset2018; Krook Reference Krook2017, Reference Krook2018; Krook and Restrepo Sanín Reference Krook and Sanín2016). Finally, comparative research also illuminates differences among women's experience of abuse and violence. Håkansson (forthcoming) reports that Swedish women officeholders at the local level encounter marginally more violence than men. But a substantial gender gap exists among higher-level officeholders than lower-level ones. Specifically, the risk of violence increases with the level of power for both women and men, but more dramatically so for women: the higher the level of power, the greater the gender gap in exposure. Additionally, women officeholders in traditionalistic political cultures are more likely than their counterparts to experience these behaviors.
Research pertaining to abuse and violence of officeholders via social media is also gaining increasing attention. These avenues of communication are used as a means of perpetrating psychological abuse and oxygenating physical violence. Together, these studies make clear that although politicians generally face expressions of anger by the public, women are subject to different types and levels of anger that often appear to be about their gender.
Although politicians praise social media as providing them with the ability to “drive their own narratives” (Di Meco Reference Di Meco2019), negatives, particularly for women officeholders, are common. For example, Barboni and Brooks (Reference Barboni and Brooks2018) report that women in three nations faced greater volumes of conversation (more comments) than men and greater attention to their appearance and relationship status, both of which were more negative than comments about men. Women were also three times more likely to see derogatory comments about their gender in social media. An examination of gender differences in tweets received by members of Parliament and candidates in the United Kingdom indicates that women were more likely than men to receive certain types of negative tweets such as hate speech (Collignon and Rudig Reference Collignon and Rudig2020; Greenwood et al. Reference Greenwood, Bakir, Gorrell, Song, Roberts and Bontcheva2019; McLoughlin and Ward Reference McLoughlin and Ward2017; Southern and Harmer Reference Southern and Harmer2019). Barboni and Brooks (Reference Barboni and Brooks2018) sum up and contextualize this body of work by concluding that women are delegitimized and depersonalized as leaders and that social media negativity can distract women leaders, instill fear in them, and dissuade them from engaging in political debate or running for office.
VIOLATING GENDER NORMS AND DISRUPTING THE STATUS QUO
Theories explaining gender differences in politicians’ experiences of psychological abuse and physical violence converge on tacit or explicit understandings of who belongs in politics, how they behave in office, and how those traditionally in power respond to newcomers to their domains. These may be termed status quo threats and gender norm violations.
Status quo threats refer to the fact that women's presence, behavior, and policy priorities shatter men's traditional preserves. Women politicians who pursue political agendas to further the status of women offend not only by holding office but also by seeking goals that may threaten men's control over offices of state. In addition to supplanting men, a spur to abuse/violence against female officeholders may be their policy differences from men. A large body of research over decades indicates that women who hold elective office in the United States, including state legislators, have distinctive policy priorities, particularly with respect to women's issues (Barnello and Bratton Reference Barnello and Bratton2007; Bratton and Haynie Reference Bratton and Haynie1999; Dodson Reference Dodson, Thomas and Wilcox1998, Reference Dodson and Carroll2001; Dodson and Carroll Reference Dodson and Carroll1991; Frederick Reference Frederick2011; Thomas Reference Thomas1994, Reference Thomas2002; Vega and Firestone Reference Vega and Firestone1995).
Regarding gender norm violations, Eagly and Karau's (Reference Eagly and Karau2002) foundational research on role congruity emphasizes that when people take on social roles that conform to the stereotypes of those roles, they are accepted. But when people take on roles that defy stereotypes or break norms, such as women who are strong leaders, they are perceived unfavorably. More explicitly, violations of public/private divisions of labor that foreground women's strengths as nurturers and men's strengths as agents may result in perceptions that women who hold political power are threats. Brescoll, Okimoto, and Vial (Reference Brescoll, Okimoto and Vial2018, 147) note that many people not only expect women and men to behave in gendered ways but believe that they “ought” to, and when they do not, moral outrage may result (see also Brescoll Reference Brescoll2011; Okimoto and Brescoll Reference Okimoto and Brescoll2010). Illustratively, Okimoto and Brescoll (Reference Okimoto and Brescoll2010) report that women candidates who are perceived to be power seeking or who exhibit power-seeking behaviors receive fewer votes than others. Such challenges to the status quo may result in efforts to discourage women in office from pursuing their full agendas and prevent others from joining (Krook Reference Krook2017).
Indeed, in the United States, there is considerable evidence that female officeholders have been and still are subject to tactics aimed at marginalizing them and decreasing their effectiveness—that is, tactics motivated by their presence. Hence, those traditionally in power may face backlash including “ridicule, condemnation, ostracism, censure” (Mansbridge and Shames Reference Mansbridge and Shames2008, 625–26). As a result, violent and abusive behaviors toward women may become more common or more extreme.Footnote 3 Similarly, Brescoll, Okimoto, and Vial (Reference Brescoll, Okimoto and Vial2018, 147) assert that women in power are likely to “incur backlash because they threaten the legitimacy of the gender hierarchy.”
We explore these theories in this research, which investigates psychological abuse and physical violence from members of the public rather than from institutional colleagues.
EXPECTATIONS
Based on theories pertaining to status quo threats and gender norm violations as well as the results of empirical research reported earlier, we expect that women state senators in our survey will report greater degrees of psychological abuse, physical violence, and sexualized abuse and violence than men senators (see Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019).
In addition, we expect that the factors correlated with experiences of abuse and violence may differ by gender as well as within groups of women. For example, it may be that holding leadership roles, such as committee chair positions or party leadership posts, affects the level of abuse and violence faced by women senators (Håkansson, Reference Håkanssonforthcoming). It may also be that age affects women and men differently, with younger women facing more abuse and violence than older women (see Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). At the heart of our expectations that women's positionality vis-à-vis political power or age affects the extent and degree of abuse and violence is the theory of intersectionality or “intersecting axes of discrimination” (Kuperberg Reference Kuperberg2018, 687). That is, women officeholders may face abuse and violence for being female in a traditionally male world as well as based on age, positional power, race, sexual identity, and other locations of power imbalance. Ideally, we would test the effects of a wide array of intersectional identities on the likelihood of experiences of abuse and violence. Yet both the universe of state senators as well as senators in our sample are not sufficiently diverse to offer meaningful quantitative analysis of each relevant identity. We are able, though, to focus on both positional power and age.
Empirical findings suggesting that power may affect experiences of psychological abuse and physical violence among women state senators come from a study of women officeholders in local politics in Sweden. Håkansson (forthcoming) reports that the risk of violence increases with the level of power for both women and men, but dramatically more so for women: the higher the level of power, the greater the gender gap in exposure. As the author notes, “Positions higher in the political hierarchy demand more assertiveness and confidence and are more associated with power-seeking behaviour than lower-level positions. Moreover, if VAWA [violence against women in politics] is driven by a dislike of female politicians, women at higher positions of power are more visible and hence more likely to be recognised as being female politicians.” In the United States, strong women mayors—those with more power in the forms of vetoes and appointments authority—reported more abuse and violence than their counterparts (Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019).
Empirical findings also indicate that younger women officeholders are more likely than their counterparts to face abuse and violence. Among the general U.S. workforce and government workforces, younger workers are more likely to face such behaviors than older employees (Piquero et al. Reference Piquero, Piquero, Craig and Clipper2013; Samnani and Singh Reference Samnani and Singh2012; Schat, Frone, and Kelloway Reference Schat, Frone, Kevin Kelloway, Kevin Kelloway, Barling and Hurrell2006). More specific to this study, Herrick et al. (Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019) and Thomas et al. (Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019) found that younger mayors were more likely to suffer psychological abuse compared to their counterparts.
Finally, when comparing our results on state senators to previous work in the United States on mayors, we expect to find more attenuated gender differences in abuse and violence among senators than mayors. The reasons for this expectation are that constituents are much less familiar with their state senator than with their mayor (see Delli Carpini and Keeter Reference Delli Carpini and Keeter1996), so displeasure about their actions may be less common; mayors are executives, and gender norm violation theory suggests that role incongruity distress is higher for executive than for legislative functions (see Eagly and Karau Reference Eagly and Karau2002); and the mayoral survey pertained to abuse and violence over the course of careers. Our state senate survey asked about experiences during the first six months of 2019. Thus, if the gender gap in abuse and violence was stronger in the past, the mayoral survey would be more likely to detect it since it asked about abuse and violence over mayors’ careers.
METHODS
Operational Definitions and Derivation of Survey Tool
Of the multiple empirical operationalizations of abuse and violence against women in politics (see Bardall, Bjarnegård, and Piscopo Reference Bardall, Bjarnegård and Piscopo2019; Biroli Reference Biroli2018; Krook Reference Krook2017; Krook and Restrepo Sanín Reference Krook and Sanín2020), we chose to pattern our categories and definitions on a 2016 IPU study and a study of U.S. mayors (Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). We do so because consistency in definitions permits comparisons across studies and populations.
Hence, our key concept are operationalized as follows:
• Physical violence: experiences of “minor” violence, such as being slapped, pushed, or subject to projectiles; significant violence, such as being shot, assaulted, or otherwise injured; and violence against property.
• Psychological abuse: exposure to insistent and uninvited behavior, attention, or verbal contact; awareness of content that was untrue and offensive in traditional media, on social media, or at public meetings; threats of death, rape, beating, abduction, or similar acts; or having a family member who received threats of death, rape, beating, abduction, or similar acts.
• Sexualized abuse/violence: indications that any of the negative experiences reported were sexual in nature.
The Survey
To focus on abuse and violence against women officeholders and to provide comparability across the few studies of this phenomenon, we modeled our survey on two earlier instruments. The first is a 2016 survey used by the Inter-Parliamentary Union to learn about female parliamentarians across the globe. The second is a survey of women and men mayors in the United States used in Herrick et al. (Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019) and Thomas et al. (Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). From July to September 2019, we conducted a survey of all state senators in the United States.Footnote 4 Survey questions explored the frequency of such experiences perpetuated by the public during the first six months of the year. We chose this operationalization so that the time frame was the same for all senatorial respondents rather than asking about such experiences during the course of an entire political career (on mayors, see Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). We also selected a time frame in which senators would have been in session. See Appendix A in the supplementary material online for survey specifics.
The survey is mixed mode with an internet version and a mail version.Footnote 5 Using a modified tailored design method (Dillman Reference Dillman2007), there were up to six contacts with each senator: (1) a pre-notice letter by mail; (2) the survey mailing; (3) a mailed reminder/thank you; (4) an emailed replacement survey; and (5) a final email reminder. Finally, the survey was mailed again to senators whose survey was originally sent to the state capitol instead of their district office. The decision to add this additional follow-up was based on evidence that surveys sent to district addresses were more likely to be have been completed.
Additionally, we sent six separate communications to state senators in part to mitigate response bias. For example, those who experienced abuse and violence might have been more inclined to respond since they had something to report, or, conversely, they may have been less likely to respond to avoid feeling or appearing to have been victimized. Those who had nothing to report may have passed on a response because of feelings that they were not adding much to the data collection. A final effort to mitigate response bias was this language in the cover letter attached to our second survey mailing: “We are interested in your experiences with the public and your colleagues whether or not they have been civil.”
A total of 252 senators out of 1,940 responded to the survey,Footnote 6 which resulted in a 13% response rate. The rate is comparable to or even larger than those of many recent studies of state legislators (Hanania Reference Hanania2017; Nownes and Freeman Reference Nownes and Freeman2019; Purtle et al. Reference Purtle, Lê-Scherban, Wang, Brown and Chilton2019). Nevertheless, we checked for the representativeness of the respondents on several traits of the full senatorial population including political party, gender, tenure in office, level of professionalism of the legislature, and crime rates in the states. The only significant relationship between response rates and these variables was with professionalism of legislatures: senators in less professional legislatures were more likely to respond.Footnote 7 Central to our analysis, 27.4% of respondents were women compared to 25.5% of nonrespondents.
Dependent Variables
The dependent variables of our study measured experiences of abuse and violence of state senators. The first question on the survey asked, “Below are several types of experiences that politicians have reported. … indicate how often you experienced each type of event from the public in the first six months of 2019 in your capacity as a Senator.” The events included the list of types of psychological abuse and physical violence. Senators were asked to indicate whether they had had these experiences never, less than monthly, once or twice a month, three or four times a month, and more than four times a month. Responses were coded on a scale from 1 to 5.
The coding we used for each of these three dependent variablesFootnote 8 is as follows: first, since few senators reported physical violence, we dichotomized this variable and coded it 1 if it was reported and 0 if not. The mean was .10 and the standard deviation was .30. The psychological abuse variable was an index that added the scores from the following variables: harassment, social media abuse, traditional media abuse, abuse at a public meeting, threats, and threats to family. Since each of these variables ranged from 0 to 4, the highest possible score was 24; the highest score for any senator, however, was 21. The lowest score was 0 (the mean equaled 4.4 and the standard deviation was 4.0). A high score meant that a senator experienced more psychological abuse than senator who had a low score. However, since the index was additive, the score for each senator does not indicate whether one type of abuse was more or less common than others. Finally, senators were asked to indicate whether any of these experiences were sexual in nature. This variable was coded 1 if yes and 0 if no.Footnote 9
Independent Variables
The key independent variable is gender (women were coded 1, men were coded 0).
We used several control variables that the literature suggests may be related to experiences of abuse and violence by officeholders in the United States (Herrick et al. Reference Herrick, Thomas, Franklin, Godwin, Gnabasik and Schroedel2019; Thomas et al. Reference Thomas, Herrick, Franklin, Godwin, Gnabasik and Schroedel2019). Hence, we included age, party, tenure, population of constituency, leadership positions held (standing committee chair or party leadership position), and education level of constituents.Footnote 10 We also included a control for professionalism of the legislature. Finally, we examined correlations among the control variables to ensure that multicollinearity was not present. Appendix B reports the means and standard deviations for each independent variable.
Models
First, we examined bivariate differences between women and men in experiences of each type of abuse and violence, the physical violence variable, the psychological abuse index, and the sexualized abuse and violence variable. Second, we used multivariate models to examine psychological abuse, physical violence, and sexualized abuse and violence while holding other variables constant. Third, we used multivariate models to examine gender differences in correlates of abuse and violence. With the multivariate analysis, we use ordered logistic regression for the psychological abuse index and bivariate logistic regression for the measure of physical violence and sexualized abuse and violence. For each of these analyses, we clustered errors by state. We report odds ratios since they are easier to interpret than maximum likelihood estimation coefficients. Finally, we weighted the data by professionalism since, as noted earlier, the only significant difference in response rate on these variables was for professionalized legislatures.Footnote 11 However, the conclusions are the same with unweighted data unless noted.
RESULTS
In this section, we report the results of bivariate and multivariate analysis for the full population of senators as well as corresponding results pertaining to the interactions of gender and each of the independent variables in our models.
In the first six months of 2019, the average state senator experienced harassment and social media abuse monthly. Approximately 60% of senators encountered harassment, 78% faced social media abuse, 55% faced abuse by traditional media, and 47% suffered abuse at a public meeting. Using the psychological abuse index, we found that 84% of state senators reported psychological abuse. Based on the physical violence measure, 10% of senators reported physical violence. Finally, 10% of all senators experienced abuse and violence that was sexual in nature.Footnote 12
Table 1 depicts the mean scores for all state senators as well as for women and men senators for each type of abuse and violence, the physical violence variable, psychological abuse index, and the sexualized abuse and violence measure. Central to our analysis, women senators reported more abuse and violence than men. For women, harassment, major physical violence, overall physical violence, and sexualized abuse and violence were statistically significant. The harassment finding is consistent with our expectation that women senators would face more psychological abuse than men. The lack of gender differences for other psychological abuse variables is at variance with our expectations. The physical violence and sexualized abuse and violence findings are consistent with expectations, however.
Psychological abuse and physical violence among U.S. state senators
Notes: The individual types of abuse and violence are coded 1–5. The psychological abuse variable is an index that ranges 0–21. The physical violence and sexualized abuse and violence variables are dichotomous. Numbers in parentheses are standard deviations. Numbers in bold are the percentages of senators who did not report any abuse or violence on the related variable. These categories are not mutually exclusive. It is possible that one incident could be included in more than on category. For example, if someone was threatened on social media, it could be classified as a threat as well as social media abuse.
* p < .10; ** p < .05.
Our multivariate model is reported in Table 2. In this table, we controlled for personal traits of state senators, districts traits, state characteristics, and legislative traits.
Multivariate analysis of psychological abuse, physical violence, and sexualized violence and abuse among state senators: Odds ratios
Notes: The psychological abuse variable is an index that ranges 0–21. The physical violence and sexualized abuse and violence variables are dichotomous. Women is coded 1 for women, 0 for men. Party: 1 = Democrat, 0 = Republican. Age is in years. Tenure is years as senator. Percent college is in whole number percentages. Population size is in actual population size in thousands. Professionalism is based on NCSL (2017) measure of full-time legislatures; it runs from 1 to 5. Numbers in parentheses are standard errors. Odds ratios are reported. Data are weighted by professionalism since the only significant difference in response rate on these variables was for professionalized legislatures. Without weighting the data tenure is significantly related to psychological abuse at the .06 level.
+ p < .10; * p < .05; ** p < .01; *** p < .001.
# Tenure is statistically significant when the data are not weighted.
First, these data show that, although women faced more psychological abuse and sexualized violence than men, the differences were not statistically significant once controls were added. Women senators were, however, more likely to have reported suffering physical violence than men (odds ratio of 2.79). In all, the multivariate results on the relationship between reports of abuse/violence and gender offer mixed support for our expectations.
Next, our multivariate model shows how experiences of abuse and violence were related to other characteristics of interest in the full population of senate respondents. First, as prior literature on general workforce and officeholder abuse and violence indicates, younger senators were more likely to have experienced abuse, violence, and sexualized abuse and violence. Party mattered as well: Democrats faced more sexualized abuse and violence than Republicans, and Republicans encountered more psychological abuse than Democrats. There were no party differences for physical violence. Third, senators who had served in office longer were somewhat more likely to have reported psychological abuse and physical violence, but they were less likely to have reported sexualized abuse and violence. Fourth, education levels of constituents were associated with psychological abuse with higher levels of education related to higher levels of abuse. However, education levels had no significant relationship with physical violence or sexualized abuse. Fifth, the larger the population of senatorial districts, the more likely senators were to encounter sexualized abuse and violence. Finally, neither the level of legislative professionalism nor holding leadership positions was significantly associated with reports of abuse or violence. Foreshadowing the next set of analyses, we also found that women senators were more likely than men to come from districts with higher levels of college education (35% to 29%), more likely to be Democrats (69% to 35%), more likely to be younger (59 years old to 63), and more likely to have shorter tenures (5 to 8 years).
To explore whether the correlates of abuse and violence varied by gender, we analyzed psychological abuse, physical violence, and sexualized abuse and violence separately for men and women. Tables 3a and 3b show several differences in the correlates of such experiences between women and men senators.
Multivariate analysis of psychological abuse, physical violence, and sexualized violence and abuse: Women
Notes: The psychological abuse variable is an index that ranges 0–21. The physical violence and sexualized abuse and violence variables are dichotomous. Women is coded 1 for women, 0 for men. Party: 1 = Democrat, 0 = Republican. Age is in years. Tenure is years as senator. Percent college is in whole number percentages. Population size is in actual population size in thousands. Professionalism is based on NCSL (2017) measure of full-time legislatures; it runs from 1 to 5. Numbers in parentheses are standard errors. Odds ratios are reported. Data are weighted by professionalism since the only significant difference in response rate on these variables was for professionalized legislatures.
+ p < .10; * p <.05; ** p < .01; *** p < .001.
# Without weighting the data, tenure is statistically significantly related to physical violence at the .06 level.
Multivariate analysis of psychological abuse, physical violence, and sexualized violence and abuse: Men
Notes: The psychological abuse variable is an index that ranges 0–21. The physical violence and sexualized abuse and violence variables are dichotomous. Women is coded 1 for women, 0 for men. Party: 1 = Democrat, 0 = Republican. Age is in years. Tenure is years as senator. Percent college is in whole number percentages. Population size is in actual population size in thousands. Professionalism is based on NCSL (2017) measure of full-time legislatures; it runs from 1 to 5. Numbers in parentheses are standard errors. Odds ratios are reported. Data are weighted by professionalism since the only significant difference in response rate on these variables was for professionalized legislatures.
* p < .10; ** p <.05; *** p < .01.
# Without weighting the data, party is statistically significantly related to sexualized violence at the .05 level and professionalism is statistically significantly related to sexualized violence at the .08 level.
The first notable finding is that the correlates of all three types of abuse and violence varied considerably between women and men senators. To start, two senatorial traits show large and important differences in women and men's experiences. The first trait concerns the effect of being a senate leader (party leader or committee chair). For men, all three models indicate that a leadership position was not statistically significantly related to psychological abuse, physical violence, or sexualized abuse and violence. In contrast, for women senators, holding a leadership position significantly increased the odds of psychological abuse and sexualized abuse and violence.Footnote 13 That women leaders experienced higher levels than other women is consistent with our expectations as well as with status quo threats, gender norm violation theories, and theories of intersectionality. Indeed, Håkansson (forthcoming) found that women officeholders in Sweden run a higher risk of political violence than men when gender interacts with level of power. That is, women experienced substantially more abuse and violence when they held higher positions.
Second, the political party of senators also made a notable difference in experiences of abuse and violence among women and men—although the nature of the relationship differed. Republican men were more likely to have reported psychological abuse than their Democratic men counterparts whereas Democratic women were significantly more likely than Republican women to have reported sexualized abuse and violence. Indeed, only one Republican woman (out of 20) experienced sexualized violence compared to 27% of the 45 Democratic women.Footnote 14 That it was mostly Democratic women who reported sexualized violence may mean that liberal women are particular targets of sexualized abuse and violence. Indeed, Krook (Reference Krook2017) suggests that women who speak from a feminist perspective are more likely to experience violence.
With respect to relationships between abuse, violence, and senatorial traits such as age and tenure in office, Tables 3a and 3b show some similarities and some differences. For women, age was significantly related to experiences of all three types of abuse and violence—with younger senators more likely to report such encounters than older women senators. However, for men, age was significant with regard to psychological abuse only—with younger men senators facing more abuse than older ones. It appears that for psychological abuse, the relationship between age and abuse/violence was direct for women and mediated for men. That is, it is possible that those who perpetrate abuse and violence feel less constrained when inflicting abuse on younger women officeholders.
Tenure in office does not significantly affect women's experiences, but among men senators, those with fewer years in office were more likely to have reported psychological abuse but less abuse and violence of a sexualized nature.Footnote 15 This may mean that for men, level of experience in office is a better predictor of abuse and violence than personal traits, and for women, what they do matters less than who they are.
Tables 3a and 3b also show that the effects of institutional context vary by gender. The findings indicate that women senators in professionalized legislatures were significantly less likely than women in nonprofessionalized chambers to have reported sexualized abuse and violence but were more likely to report physical violence. In contrast, professionalism was not significantly related to the experiences of men on any of the three measures.Footnote 16 Another gender-differentiated finding is that larger population sizes of senatorial districts were significantly correlated with less psychological abuse for women; population sizes were not significantly correlated to any of the three dependent variables. Finally, only district educational levels had similar effects for women and men state senators. For both, higher educational levels were significantly correlated to experience of psychological abuse. A possible explanation for the fact that only one variable, educational levels, affected women and men state senators’ encounters of psychological abuse similarly is that so much abuse is perpetuated via social media—and the facility to use social media depends, in part, on education—and the resources associated with higher levels of education.
In sum, there are several differences in the correlates of women and men's experiences of psychological abuse and physical violence. Most notable for this study, women state senators who held leadership positions were more likely than other women to face abuse and violence. The relationship was not present for men senators. Additionally, lower levels of legislative professionalism were inversely correlated with abuse and violence for women, but not men. These findings suggest that the expression or forms of abuse and violence for women may differ from those of men. Therefore, our findings are consistent with our expectations and findings from empirical research. Moreover, they are consistent with theories of status quo threats, gender norm violations, and intersectionality.
DISCUSSION
In this study, we report results of a survey of U.S. state senators’ experiences of psychological abuse, physical violence, and sexualized abuse and violence, as well as gender differences among senators.
Overall, in the first six months of 2019, state senators in the United States reported encountering meaningful psychological abuse and physical violence. In all, 84% faced psychological abuse, 10% suffered physical violence, and 10% reported sexualized abuse and violence. Bivariate analysis shows that, compared to men, women suffered more of four specific types of abuse and violence: harassment, major physical violence, overall physical violence, and sexualized abuse and violence. In contrast, multivariate analysis shows that the only statistically significant gender difference in encounters of abuse and violence was that women senators faced more physical violence than men.
Critically, we also found that different factors contributed to abuse and violence against women and men senators. Most notably, women with higher levels of power (party or committee leaders) were more likely than other women to experience psychological abuse and sexualized abuse and violence; women in more professionalized legislatures reported more physical abuse and less sexualized abuse and violence than their counterparts; younger women faced more of all three types than older women senators; and Democratic women senators faced more sexualized abuse and violence than Republican women. These results comport with theories of gender norms, status quo violations, and intersectionality: when women behave in ways that are inconsistent with traditional expectations, they may face negative consequences. And Democratic women who are considered to be more liberal, and therefore greater violators of gender norms, may face more behavior intended to restore the status quo.
At the outset of our study, we also expected that, compared to mayors, the only other group of U.S. officeholders for whom data are available, senatorial gender differences would not be as robust. That expectation was confirmed. The reasons we expected attenuated results include that constituents are much less familiar with their state senator than with their mayor. Thus, contact with senators and displeasure or anger toward their actions may be less common. Second, mayors are executives, and gender norm violation theory suggests that role incongruity distress is higher for the executive than the legislative function. Third, the mayoral survey pertained to episodes over the course of careers. Our state senate survey asked about experiences during the first six months of 2019. Thus, if the gender gap in abuse and violence was stronger in the past, the mayoral survey would be more likely to pick that up.
Implications
Evidence of psychological abuse and physical violence among state senators may weaken democracy by depressing the pool of public servants. Concerns about physical safety and psychological health of officeholders and their families may impel some serving officers to attenuate their public service and discourage others from running for office. Our findings also raise questions about increasing gender diversity in office. It is possible that more women than men may avoid running for office because they are likely to experience more and more types of abuse and violence. Coupled with research suggesting that women state legislators are less likely than men colleagues to seek higher office (Einstein and Glick Reference Einstein and Glick2017; Fulton et al. Reference Fulton, Maestas, Sandy Maisel and Stone2006; Maestas et al. Reference Maestas, Fulton, Maisel and Stone2006) and that they are more affected than men by the high costs of running for and service in elective office (Lawless Reference Lawless2012; Shames Reference Shames2017), women's underrepresentation may not be alleviated in the near future. This matters for reasons of descriptive and substantive representation.
Another implication is that it may be more difficult for senators to be effective under the stress of abuse and violence. Such incidents may be upsetting and distracting and, thereby take time and focus away from their representative duties. Politicians who experience abuse and violence often suffer psychologically from the attacks (Every-Palmer, Barry-Walsh, and Pathé Reference Every-Palmer, Barry-Walsh and Pathé2015; Herrick and Franklin Reference Herrick and Franklin2019).
Strengths and Limitations of Our Research
In addition to illuminating the results of this study of U.S. state senators, it is important to note limitations.
A first set of limitations pertains to response rates and patterns. Although the response rate to our survey was higher than other recent surveys of the same population, it was low. Relatedly, our sample sizes were not large enough to uncover differences among many groups of women, especially pertaining to race and ethnicity. This is a particular concern as minority women may be especially likely to encounter violence (IPU 2016; Kuperberg Reference Kuperberg2018). It is also possible that self-selection bias could have skewed the results. More state senators who experienced abuse and violence than did not may have responded—or the reverse. Similarly, our results may be affected by those who left their positions prior to the survey because of abuse and violence. These are common issues with survey research that future researchers might mitigate by exploring the same questions with complementary methodologies. Such approaches may include in-depth interviews and analysis of news reports or official complaints.
Another set of limitations pertains to the scope of political abuse and violence toward officeholders and their gendered nature. First, we did not examine violence against officeholders from colleagues—other officeholders, staff, the media, or lobbyists—which would address theories of gendered institutional challenges. Second, we did not explore issues of economic violence or semiotic/symbolic violence. Third, our survey does not shed light on abuse and violence against the women resulting from a desire to keep women out of office. Finally, because we surveyed officeholders, we know little about the perpetrators of the abuse and violence.
In all, this research is an important step toward increasing empirical evidence of psychological abuse and physical violence among women and men in elected office. More work is needed to deepen our knowledge of gender differences in types, amounts, and correlates of these behaviors. The quality and diversity of our representation depends on it.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S1743923X2000063X
