Sexual Harassment, Workplace Authority, and the Paradox of Power

Sexual harassment is classified as a form of sex discrimination under Title VII of the Civil Rights Act of 1964. The U.S. Equal Employment Opportunity Commission (EEOC) defines it as “unwelcome sexual advances, requests for sexual favors, and other verbal or physical conduct of a sexual nature” that interferes with one’s employment or work performance or creates a “hostile or offensive work environment” (U.S. EEOC 2011). Due, in part, to varying definitions and indicators, prevalence estimates vary dramatically (Welsh 1999), leading many researchers to adopt a strategy of triangulation that considers multiple forms or measures (e.g., Houston and Hwang 1996; Uggen and Blackstone 2004).

Feminist scholarship situates sexual harassment within broader patterns of discrimination, power, and privilege, linking harassment to sex-based inequality (MacKinnon 1979). Quinn’s (2002) research on “girl watching,” for example, ties patriarchal gender relations to everyday workplace interactions. Quinn argues that other men, rather than women, are often the intended audience of sexist gestures and comments. Although men often view “girl watching” as light-hearted and playful, and seem surprised when women take offense, such activities demonstrate men’s power to sexually evaluate women.¹ Similarly, Martin (2001) finds that men “mobilize masculinities” in ways that often exclude and cause harm to women as a group, even when this is not their intention.

Connell’s (1987) theory of hegemonic masculinity, which argues that society privileges a single normative ideal of male behavior, provides a broad sociological framework for understanding harassment, gender, and power. Men may be vulnerable to harassment if they are perceived as feminine (DeSouza and Solberg 2004; Waldo, Berdahl, and Fitzgerald 1998), and women may be targeted if they challenge their subordinate position in the gender system. Sexual harassment may thus act as a tool to police appropriate ways of “doing gender” in the workplace and to penalize gender nonconformity (West and Zimmerman 1987).

Research on contrapower harassment suggests that gender, race, and class positions imbue harassers with informal power, even when targets possess greater organizational authority than do their harassers (Rospenda et al. 1998). Women holding authority positions thus offer an intriguing paradox for theory and research on sexual harassment, and scholars have advanced two distinct positions. The first, the vulnerable-victim hypothesis, suggests that more vulnerable workers—including women, racial minorities, and those with the most precarious positions and least workplace authority—are subject to greater harassment. The second, the power-threat model, suggests that women who threaten men’s dominance are more frequent targets. Although the matter is far from settled, research has found greater support for the paradoxical power-threat model, in which women in authority positions are most likely to face harassment (Chamberlain et al. 2008) and discrimination (Stainback, Ratliff, and Roscigno 2011).

Women supervisors, who hold authority over some men, directly challenge the presumptive superiority of men. Women continue to be underrepresented in positions of authority or relegated to the lower rungs of management (Elliott and Smith 2004; Gorman 2005; Kalev 2009; Reskin 2003; Reskin and McBrier 2000). When women are able to crack the glass ceiling and attain leadership positions, stereotypical gender beliefs about their “natural” abilities continue to shape perceptions of their job performance (Davidson and Cooper 1992; Eagly and Carli 2007). Moreover, while men in traditionally female occupations reap the rewards of a glass escalator to leadership positions (Hultin 2003; Williams 1992), women supervisors are often isolated and seen as undeserving of their positions (Ridgeway and Correll 2004; Ridgeway and Smith-Lovin 1999). In fact, women are unlikely to be promoted to management unless a sizeable proportion of women are already in place, highlighting the difficulty of gaining initial entry to such positions (Cohen, Broschak, and Haveman 1998). Taken together, these processes point to women supervisors as potential targets for harassment.

The idea of masculine overcompensation—in which men react to threats to their manhood by enacting an extreme form of masculinity (Willer 2005)—also helps explain why men may harass women in power. Maass and colleagues (2003), for example, find that male participants in a computer image-sharing task sent more pornographic and offensive images to females identifying as feminists than to females adhering to more traditional gender roles. Along similar lines, Das (2009) concludes that females who are “too assertive” threaten the gender hierarchy and are denigrated through harassment. Correspondingly, De Coster, Estes, and Mueller (1999) find that females with greater tenure, independent of age, are more likely to view sexual harassment as a problem for them at work, concluding that the practice is used instrumentally against powerful females who encroach on male territory. Each of these findings suggests that women supervisors may be more likely than other working women to experience sexual harassment.

It is also possible that supervisors report greater rates of harassment simply because they are more aware of the phenomenon. Supervisors’ advanced education and training likely increase their overall legal consciousness and understanding of sexual harassment. As a result, supervisors, who are often responsible for fostering a professional work environment free from harassment and discrimination, may be more likely to recall sexualized workplace interactions and to label such experiences as harassment.

Supervisory authority and expressions of gender are also tied to other forms of sex-based discrimination (Stainback et al. 2011). Berdahl (2007a:644) reconceptualizes sexual harassment as sex-based harassment, defined as “behavior that derogates, demeans, or humiliates an individual” based on sex. Sex-based harassment is driven by a motivation to protect sex-based social standing. As a result, targets are most likely to be females who threaten males’ status. For example, Berdahl (2007b) finds that females who perform gender in stereotypically masculine ways (e.g., assertive, dominant, and independent) are more likely to experience harassment. She argues that harassers reinforce masculine dominance by relegating women to the “low status of being a means to a man’s sexual ends” (Berdahl 2007a:649). When men are targeted by other men, harassers “prove” their own manhood by undermining their targets’ masculinity.

Apart from supervisory authority and expressions of gender, workplace demography also influences harassment experiences. Sexual harassment occurs across a diverse range of job settings (Collinson and Collinson 1996; Dellinger and Williams 2002; Giuffre and Williams 1994), but a large literature debates the importance of numerical sex ratios (Welsh 1999). The weight of the evidence suggests that harassment, of both men and women, most often occurs in male-dominated work settings (Fitzgerald et al. 1997; Mansfield et al. 1991; Rospenda et al. 1998; Uggen and Blackstone 2004). Based on dynamics of sex, gender, and power in the workplace, we form three primary hypotheses:

After testing these hypotheses in our quantitative analyses, we turn to qualitative interview data to understand the underlying process behind these patterns. By probing the experiences of our survey participants, we are better able to explain why and how gender, sex, and power shape harassment experiences and workplace interactions more broadly.

We analyze longitudinal data from the Youth Development Study² (Mortimer 2003). The study began in 1988, when participants were 9th graders in St. Paul, Minnesota public schools. The sample consists of 1,010 youth, who have since been surveyed regularly. For this article, we analyze the 2003 and 2004 surveys, when participants were approximately 29 and 30 years old. Survey retention was 73 percent in 2004, and our analytic sample (N = 522) consists of all working participants who responded to sexual harassment items in both waves.

Quantitative studies are sometimes criticized for assuming that all sexuality in the workplace is harmful to women, or that women who do not label their experiences as harassment are suffering from false consciousness (Williams, Giuffre, and Dellinger 1999). Because workers may experience some sexual behavior as tolerable or pleasurable (Dellinger and Williams 2002; Giuffre and Williams 1994; Lerum 2004; Schultz 2003; Williams 1997), YDS respondents were asked about sexual behaviors that they considered offensive or that made them uncomfortable. Although this wording admits a range of conduct that may fall short of legal definitions of harassment, it clearly eliminates sexualized interactions that respondents would consider pleasurable. Finally, to avoid potential biases associated with behavioral indicators, we also asked respondents directly whether they would classify their own experiences as sexual harassment.

Relative to surveys, however, qualitative research is often “better suited to discovering how the meaning of sexual behaviors varies in different organizational contexts” (Dellinger and Williams 2002:244). Dellinger and Williams (2002), for example, powerfully contrast the “locker room” culture of a male pornographic magazine with the “dorm room” atmosphere of a feminist magazine, showing how organizational culture shapes workers’ definitions of harassment. Our survey results reveal patterns of association, but they cannot speak directly to the more subtle and specific mechanisms linking group membership to opportunities and outcomes (Gross 2009; Reskin 2003). We therefore also conducted interviews with 33 YDS respondents (14 men and 19 women) to more closely examine the processes linking power and gender and to better contextualize harassment experiences.³ Of the interviewees, all but seven identified as white, most self-identified as straight, and they reported a range of supervisory and nonsupervisory jobs. We asked interview participants a series of open-ended questions about their work histories and interactions, inviting them to share what they felt was most important based on our interest in harassment, workplace problems, and workplace sexuality. All interview participants’ names are pseudonyms.

Table 2 shows estimates predicting the likelihood of experiencing any sexually harassing behaviors (Models 1 and 2), the number of harassing behaviors experienced (Models 3 and 4), and the likelihood of subjectively defining these experiences as sexual harassment (Models 5 and 6). Although YDS respondents self-reported harassing experiences, there is great ambiguity in the meaning and interpretation of these events. The first four models identify individuals most likely to experience harassing behaviors, but they do not tell us whether targets themselves considered this to be sexual harassment. Over one-third of our sample reported at least one behavioral indicator of harassment in 2004, yet only 7 percent subjectively reported experiencing sexual harassment (11 percent of females but only 1 percent of males). The final two models in Table 2 concern subjective harassment, which represents a higher threshold than the behavioral measures. Here, we can be confident that reports do not reflect desired sexual banter or behavior that was reciprocated by targets. Because the process of defining and labeling behaviors as sexual harassment varies by sex (Uggen and Blackstone 2004) and few males reported experiencing harassment, we focus this analysis on females only. We also exclude nativity from these models because the small number of female immigrants led to unstable estimates.

Consistent with our first hypothesis, supervisory authority has significantly different effects for males and females. Under the interaction coding in Models 1, 2, 3, and 4, the main supervisor and femininity coefficients indicate effects for females, and the supervisor × male and femininity × male interactions show the difference in effects for males and females. In all models, workplace power is a significant predictor of harassment for females. Even with the strong control for past harassment, female supervisors are 138 percent more likely to experience any harassing behaviors (e^.868 = 2.38 in Model 2), they report a rate of harassment 73 percent greater than that of nonsupervisors (e^.547 = 1.73 in Model 4), and they are nearly three and a half times as likely to subjectively interpret these experiences as sexual harassment (e^1.245 = 3.47 in Model 6). The interaction of sex and supervisory authority (supervisor × male) is significant in Models 1 and 2 but falls just short of standard significance levels in Models 3 (p = .070) and 4 (p = .066). Taken together, these results suggest that not only are female supervisors more likely to experience harassing behaviors (Models 1 and 2) and to label them as such (Models 5 and 6), but they also report a more varied and sustained form of harassment (Models 3 and 4).¹⁰ See Appendix Table A1 for a descriptive comparison of females and males by supervisory status.

We find mixed support for our second hypothesis regarding gender conformity. In Models 1, 2, and 3, more feminine females are less likely to report harassment. The femininity × male interaction also signals a significant sex difference in the effect of femininity. Addition of the lagged measure in Model 4, however, reduces the estimated effects below standard significance levels.¹¹ This pattern could indicate a spurious relationship in Model 3 or it could be indicative of statistical overcontrol in the lagged models. Either way, the effect of femininity and its interaction with sex is less robust than the effect of authority position and its interaction with sex.

Industry sex ratio, however, emerges as a significant predictor of subjective harassment in Models 5 and 6, lending support to our third hypothesis. In the trimmed models reported in Table S1 of the online supplement, industry sex composition is also a significant predictor of the number of harassing behaviors. These findings may reflect, in part, the different nature of harassing behaviors in male-dominated work settings. As suggested by our interview data (see below), behaviors such as suggestive stories or inappropriate comments about women’s bodies may be interpreted as more menacing, malicious, or degrading at jobsites where women are socially and numerically isolated. Just as industry sex ratio is predictive only of subjective harassment, it is also plausible that when women are surrounded by men, they may interpret behaviors differently and be more likely to label sexualized behaviors as harassment than will women in more gender-balanced settings.

Our statistical control for whether respondents experienced sexual harassment in the previous year is a strong and significant predictor in all models, which likely captures stable individual differences in legal consciousness as well as stability in targeting. Respondents who reported any harassing behaviors in 2003 were over six and a half times as likely to report harassing behaviors in 2004 (e^1.887 = 6.60 in Model 2); respondents who were subjectively harassed in 2003 were seven times as likely to report subjective harassment the following year (e^1.944 = 6.99 in Model 6). This lagged measure is perhaps most important in models predicting subjective harassment, because the lag effectively controls for stable within-person characteristics that may influence whether a person interprets an event as sexual harassment or dismisses it as sexual banter.

Moving to individual characteristics, we observe little difference in the likelihood of reporting any harassing behaviors by race, although the rate of harassment among whites is lower than for persons of color in Models 3 and 4. Respondents born in the United States are also more likely to experience harassment in Model 1, although this effect diminishes with the introduction of the lagged dependent variable. In separate analyses, we examined how sex, supervisory authority, and race intersect to shape harassment experiences (see Table S3 of the online supplement). Among white females, 47 percent of supervisors and 29 percent of nonsupervisors reported at least one harassing behavior (p < .01). Workplace power does not significantly increase the odds of harassment for females of color, but it does not serve as a protective factor either: 43 percent of supervisors and 45 percent of nonsupervisors experienced harassing behaviors. We observe no significant differences among males. These results suggest that harassment is not a unidimensional experience, shaped by just one aspect of a person’s identity. Instead, as demonstrated by intersectionality scholars such as Crenshaw (1991) and Collins (2004), these dimensions operate simultaneously to produce inequalities.

Our quantitative analyses establish the basic empirical relations, but we turn to qualitative interviews with YDS participants to delineate why and how these relations matter. To learn more about the processes connecting gender and power, we present respondents’ accounts of harassment in the context of our three hypotheses: (1) experiences of women in positions of authority; (2) experiences of workers who do not conform to culturally dominant expectations of gender; and (3) patterns of harassment in male-dominated industries and occupations.

This multimethod analysis offers the strongest evidence to date on the interaction of sex, gender, and power in predicting sexual harassment (Berdahl 2007a; Connell 1987; Quinn 2002; Uggen and Blackstone 2004). The vulnerable victims perspective suggests that authority acts as a protective factor, exempting women from the suggestive gaze or unwelcome touch of co-workers, but we find that supervisory status actually increases women’s harassment, in keeping with the power threat perspective. Although other research has suggested this counterintuitive result, this study is the first longitudinal investigation to clearly reveal the pattern.

Could this finding be a product of supervisors’ greater sensitivity or legal consciousness? If so, we would expect to have observed stronger effects of supervisory status on subjective harassment (in Models 5 and 6 of Table 2) than on behavioral indicators (in Models 1, 2, 3, and 4). By every measure, however, female supervisors were more likely to report harassment. Moreover, although supervisors may have greater awareness of sexual harassment law and policy, most individual differences in consciousness should be absorbed in the lagged dependent variable measured just one year prior. Legal consciousness undoubtedly evolves over the life course, but it is unlikely that supervisors’ consciousness changed so dramatically over a 12-month period. In addition, no evidence suggests that either consciousness or the lagged dependent variable would operate differently for male and female supervisors. Instead, our survey and interview data extend and confirm research connecting sexual harassment with women’s workplace power (Berdahl 2007a; Quinn 2002; Willer 2005).

Within the supervisor category, there may be important differences in harassment by level of workplace authority. For example, women in upper management, or in organizations with few women in supervisory roles, may experience more harassing behaviors (Cohen et al. 1998). When we distinguished between females who supervised only one to three workers and those who supervised a greater number, we found larger positive effects of supervisory authority on harassment for females supervising a larger number of workers. Further research is needed, however, to move beyond a simple supervisor/nonsupervisor dichotomy and to better locate workers in hierarchies of authority. Studies of harassment in the armed forces, for example, could more clearly speak to finer-grained differences by rank (Miller 1997).

Regardless of organizational rank, sexual harassment objectifies workers and reduces women to sexual objects in ways that “may trump a woman’s formal organizational power” (Quinn 2002:392). Indeed, our qualitative interviews help to explain why women in authority positions are targeted for harassment. Women supervisors repeatedly spoke about feeling isolated and of harassment by co-workers and subordinates directed toward putting them “in their place.” Still, they tolerated such harassment to keep their jobs. Social isolation may also represent an important mechanism linking expressions of gender and industry sex ratios to harassment, in keeping with our second and third hypotheses. Whether attempting to prove they could lead a team of workers or prove themselves as women in masculine fields, women’s isolation in these positions repeatedly left them vulnerable to harassment. Women supervisors were told “this is no place for women,” and men and women who diverged even slightly from rigid gender expectations elicited taunts and more menacing responses.

Although our longitudinal approach is well-suited to testing our hypotheses, we must note several caveats. First, the YDS is based on a Minnesota sample that is not nationally representative. Second, our quantitative analysis is based on harassment occurring when respondents were 29 or 30 years old. Although analysis of a single cohort is advantageous in distinguishing authority effects from age effects, supervisory authority and gender may operate differently in other phases of the life course. Further research is needed to consider potential interactions between age, gender, and power, as well as changing definitions of harassment over time. Third, although we measured harassment behaviorally and subjectively—increasing confidence in our findings—we relied on targets’ self-reported accounts. There is a large difference between the percentage of respondents who reported harassing behaviors relative to the number who defined their experiences as sexual harassment. This discrepancy is attributable, in part, to differing definitions of harassment among respondents, a reluctance to label oneself as a target or a victim, and differences in organizational cultures and norms regarding workplace sexuality (see Dellinger and Williams 2002; Ewick and Silbey 1998). Finally, the vast majority of our YDS sample self-identified as straight. More qualitative and large-scale survey research on harassment experiences of workers expressing other sexual identities is clearly needed.

Despite these caveats, this study provides clear evidence on the effects of workplace authority on sexual harassment, with consequential implications for theories of gender and power. In particular, we find that female supervisors are more, rather than less, likely to be harassed, supporting the notion that interactions between workers are not driven strictly by organizational rank. Instead, co-workers’ relative power is also shaped by gender. Although women supervisors’ authority is legitimated by their employer, sexual harassment functions, in part, as a tool to enforce gender-appropriate behavior. Formal lines of authority are established to maximize efficiency and help organizations run more smoothly, but tensions surrounding gender or other forms of inequality will likely manifest in other ways (Roscigno 2011). When women’s power is viewed as illegitimate or easily undermined, co-workers, clients, and supervisors appear to employ harassment as an “equalizer” against women supervisors, consistent with research showing that harassment is less about sexual desire than about control and domination (Berdahl 2007a, 2007b; Schultz 2003).

Sexual harassment policies are put in place to protect workers, but organizational practice is often misaligned with formal policies or grievance procedures, calling into question fundamental assumptions many sociologists make regarding organizational constraint and agency (Roscigno 2011). Take, for instance, the sexual harassment of Holly at a company dinner. Her co-workers’ concern led to her removal from the situation rather than the punishment of her harasser. Her colleagues ended Holly’s harassment, but they did so in a way that would not jeopardize their profitable business relationship with the client. Holly’s co-workers quickly labeled her harasser as a “bad apple,” but the organizational contexts that allowed the harassment and broader gender inequality to occur remained unquestioned.

Many of our interviewees experienced wide-ranging forms of sex-based discrimination as well as sexual harassment. These findings may thus extend more broadly to workplace bullying, intimate partner violence, and other forms of discrimination and harassment. For example, Johnson (1995:284) identifies “patriarchal terrorism” as a form of intimate partner violence that reflects “patriarchal traditions of men’s right to control ‘their’ women.” Similarly, Macmillan and Gartner (1999) direct attention to dyadic power relations, finding that women’s employment lowers risks of spousal abuse when their male partners are employed but increases risks when their male partners are unemployed. Future research might also examine sexual harassment dyads and the positions and perspectives of harassers (Quinn 2002). Harassers appear to view targets as a threat to their own power—in the workplace and in the broader society.

Theory and research on gender stratification often make the implicit assumption that problems such as sexual harassment, sex discrimination, and workplace bullying will recede if and when women attain greater power at work. Yet power in the form of supervisory authority also provokes backlash from clients, subordinates, and fellow supervisors. This paradox of power represents a challenge and an opportunity for existing frameworks. Beyond gender, characteristics such as race and class may similarly trump formal organizational authority in determining workplace power. Firms are increasingly adopting policies to increase diversity in management (Kalev, Dobbin, and Kelly 2006), but this study points to a new obstacle for women and, perhaps, racial minorities in leadership positions (but see Hirsh and Lyons 2010). To test the latter idea, an extension of this analysis might examine whether racial minorities who supervise others are subject to greater racial harassment.

Although legal and organizational responses to sexual harassment have evolved rapidly in the past three decades, the cultural image of harassers and targets has not kept pace with changing workplace realities. Many still view the typical harassment scenario as one involving a sleazy male boss and a powerless female secretary. As this article shows, the reality is far more varied. Moving away from such stereotypes is a critical step for improving organizational policies and training procedures on sexual harassment. Effective training must go beyond male boss/female subordinate role-playing exercises and better reflect the diversity of harassment experiences. Effective grievance procedures must also enable targeted workers to come forward without undermining their own authority. For women who become bosses, their positions create a paradox of power in a gender system that continues to subordinate women. In taking on positions of authority, they also take on a greater risk of sexual harassment.

We thank Jeylan Mortimer for providing the data and support for this research, and Jeremy Minyard, Phyllis Moen, and ASR reviewers and editors for helpful comments on earlier versions of this manuscript.