Introduction

Research on workplace sexual harassment has a distinguished history (e.g., Berdahl, Magley, & Waldo, 1996; Fitzgerald & Shullman, 1993; Gruber, 1998; Gutek, 1985; LaRocca & Kromrey, 1999; O’Connell & Korabik, 2000; Tangri, Burt, & Johnson, 1982; Uggen & Blackston, 2004; Wayne, 2000; Williams, Giuffre, & Dellinger, 1999). Most of this research, however, has been focused on harassment of women by men, leading some scholars to call for more research on the harassment of men (Pan, 1994; Vaux, 1993). Despite such calls, there have been few empirical or nationally representative studies of the correlates and prevalence of men’s workplace sexual harassment and comparisons with women’s harassment. Using data from the nationally representative 1992 U.S. National Health and Social Life Survey (NHSLS), this study begins to fill these gaps. In addition to providing baseline data on the prevalence of workplace sexual harassment among American women and men, it examined risk factors for sexual harassment among both genders.

Theoretical Models of Sexual Harassment

While the literature on men’s sexual harassment is sparse, two major theoretical models emerge from the literature on women’s harassment: (1) the power differentials approach, and (2) a “routine activities” model derived from criminology, emphasizing risk factors emerging from a victim’s attributes or behavior and from the social context.

Power Differentials

Studies of women’s sexual harassment have largely been based on the power differentials approach, focusing on gender asymmetries in power in patriarchal societies, that may make a woman vulnerable to sexual harassment by more powerful men (Berdahl et al., 1996; Bernard & Schlaffer, 1997; Cleveland & Kerst, 1993; O’Connell & Korabik, 2000; Sheets & Braver, 1999; Tangri et al., 1982; Waldner, Vanden-Goad, & Sikka, 1999; Wilson & Thompson, 2001). Such asymmetries can stem from societal norms, organizational hierarchies, or interpersonal characteristics, and can involve formal or informal power. As has been noted elsewhere (Parish, Das, & Laumann, 2006), two possible sub-hypotheses can be derived from this approach. In the first, the vulnerable victim hypothesis, harassment is driven by a victim’s lesser power than potential perpetrators to whom she might be exposed, with power determined directly by an organization’s formal hierarchy (Tangri et al., 1982; Wilson & Thompson, 2001) or indirectly, through an organization’s culture or societal norms (O’Connell & Korabik, 2000; Tangri et al., 1982; Wilson & Thompson, 2001).

Abstracted away from its focus on women’s harassment, this “victim vulnerability” mechanism could conceivably account for men’s harassment as well, whether by women in positions of greater power, as illustrated in media reports (Armour, 2004; Gross, 1995; Lawlor, 1994) or by other men. A few speculative essays and qualitative or small-sample studies, for instance, have typically viewed male harassment as a manifestation of hostile sexism, perpetrated by dominant heterosexual men upon weaker, more effeminate or homosexual males as a means of defending traditional gender roles (DeSouza & Solberg, 2004; Fiske & Glick, 1995; Glick & Fiske, 1996, 1997). However, large-sample survey data suggest that male harassment may be perpetrated more often by women. For instance, in both the 1980 and 1994 waves of the U.S. Merit Systems Protection Board study, among those in federal government jobs, only a minority of men harassed over the past 24 months (22% of harassed men in 1980 and 21% in 1994) reported other men as perpetrators, with the proportions remaining stable over the 14 year period (DeSouza & Solberg, 2004; U.S. Merit Systems Protection Board, 1981, 1988, 1995). While the NHSLS data provide no information on perpetrators, it is reasonable to assume that men in less powerful positions would be more likely to face harassment, whether from dominant men or women.

The second, power-threat, hypothesis (under the “power differentials” model) postulates that women who become “too assertive,” thereby threatening male dominance, are denigrated through harassment. Thus, for instance, women in positions of higher authority are more likely to be victimized (De Coster, Estes, & Mueller, 1999; Mueller, De Coster, & Estes, 2001; Tangri et al., 1982). Hypotheses 1a and 1b (Table 1) reflected these conjectures about victim vulnerability and power-threat. Since the power-threat mechanism seemed unlikely to apply to men, the reference for Hypothesis 1b was solely to women’s harassment.

Table 1 Summary of competing hypotheses by type of mechanism in harassment
Full size table

Routine Activities: Perpetrator’s Perception of Benefit

A second possible model derives from research in criminology, on risk factors for crime emerging from a target’s attributes and his/her pattern of daily or routine activities (Clarke & Felson, 1993; Cohen & Felson, 1979; Cohen, Kluegel, & Land, 1981). Three core mechanisms are emphasized in this literature: increased benefit perceived by a potential perpetrator (based on a target’s attractiveness), and his/her increased opportunity (through greater contact with a victim) for and lower cost (due to lower probability and severity of sanctions) of harassment.

With regard to a perpetrator’s perception of benefit, it might be argued that women with more sexualized personality structures consciously or unconsciously send out cues that are interpreted as receptivity to sexual attention. From an evolutionary standpoint, such signaling may represent for potential partners—especially those following short-term mating strategies (Buss & Schmitt, 1993)—an apparent solution to mate-identification problems. Potentially, the same mechanism could apply to sexualized men as well. The NHSLS data contained no direct indicators of a victim’s sexual signaling. However, sexualization itself, as a precursor to signaling, could arguably stem both from a target’s developmental trajectory and from later-life experiences that get layered onto this basic personality substrate. The literature on pubertal development suggests, for instance, that early puberty is correlated with higher levels of sexuality in adulthood, whether as a simple marker of differential hormonal levels or by serving as a social stimulus (Belsky, Steinberg, & Draper, 1991; Moffit, Caspi, Belsky, & Silva, 1992; Udry, 1988). A growing number of large-sample studies on “repeat victimization” also suggest that, among both genders, early events, such as pre-pubertal sexual contact, may channel a person into more vulnerable life paths (possibly through the sexualization mechanism), with victimization becoming an enduring condition rather than an isolated event (Browning & Laumann, 1997, 2003; Finkelhor, Ormrod, & Turner, 2007; Laumann, Browning, Rijt, & Gatzeva, 2003; Parish et al., 2006; Svedin & Priebe, 2007). These potential linkages between harassment, early puberty, and early sexual contact were studied through Hypothesis 2a (Table 1).

Among later-life influences, more education might also lead to the internalization of less inhibiting “cultural scenarios” of sexual appropriateness (Ellingson, Laumann, Paik, & Mahay, 2004), whether through the educational process itself or by embedding a person in more sexually permissive peer networks. An individual’s total number of sex partners over the lifetime is arguably a good indicator of a sexualized life trajectory. Diversified sexual interests—as indicated, for instance, by the number of sexual practices one finds potentially appealing—could be a good proximal indicator of current sexualization. The same may be true with frequent masturbation, whether conceptualized as an outcome of life course processes or hormonal levels.

These conjectures led to Hypothesis 2b (Table 1).

Routine Activities: Perpetrator’s Greater Opportunity and Lower Cost

A second mechanism under the routine activities model is a victim’s heightened exposure to potential harassers in the course of his/her routine daily activities, increasing a perpetrator’s opportunity for harassment (Clarke & Felson, 1993; Cohen et al., 1981; Parish et al., 2006). For instance, previous research suggests that harassment is particularly common for women in occupations where there is greater contact with customers and co-workers (Hughes & Tadic, 1998; Parish et al., 2006), a factor that could potentially apply to men’s harassment as well. In general, women working on farms, many of which may be family owned, could face lower exposure to strangers and casual acquaintances than those in any other occupational category. Similarly, men in “office” jobs—at any level—could face more exposure to any potential harassers than farmers, and to opposite-gender harassers than those in manual labor (a male-intensive profession). A growing literature also suggests that both women and men in military service face high levels of sexual harassment, due possibly to living in proximity with others in a military base, and to a blurring of professional and personal relationships (Antecol & Cobb-Clark, 2001; Fitzgerald, Drasgow, & Magley, 1999). Since only a very small proportion of women in these data had ever served in the military, this factor was expected to apply significantly only to men. These conjectures led to Hypothesis 2c (Table 1).

Additionally, a perpetrator’s cost of harassment might be lower with victims routinely working nights or weekends, given the lower likelihood of observation and sanctions from third parties. Accordingly, Hypothesis 2d (Table 1) was included.

Background Conditions

Several additional background characteristics were controlled in the analysis, including current age, age at first job, current unemployment, ethnicity, foreign birth, and religious affiliation. Controlling for age at entry into the labor force, current age indexed “years at risk,” or cumulative exposure of an individual to the risk of harassment for a greater number of years. Participant age could also index cohort effects. In particular, if male harassment is a more recent phenomenon, then men in the younger cohorts could be more likely to have experienced lifetime harassment, with these two separate mechanisms difficult to tease apart in cross-sectional data. Additionally, especially among women, being currently out of work could be a marker of infrequent entry into the labor force and hence a shortened time at risk. In this sense, both age and current employment also indexed opportunity (under the routine activities model) in a temporal sense. Hypothesis 3a reflected these conjectures.

The sparse literature on the linkage between sexual harassment and ethnicity is ambiguous, with small-sample studies differing, for instance, on whether black women are harassed more (Gruber & Bjorn, 1982; Mansfield et al., 1991) or less (DiVasto et al., 1984; Wyatt & Riederle, 1995). In the literature (Wyatt & Riederle, 1995), fewer reports of harassment (and other sexually abusive events) among black women has been interpreted as a case of underreporting, driven by a belief that community support would be lacking upon disclosure—a factor that would arguably not apply to anonymous survey data such as the NHSLS. Next, foreign birth could imply less sensitivity to sexual attention, with individuals from societies with more liberal norms about workplace sexual attention less likely to label such behaviors as unwanted or offensive (Shupe, Cortina, Ramos, Fitzgerald, & Salisbury, 2002). Finally, both women and men with no current religious affiliations and those affiliated with non-conservative religious groups could perhaps experience (through more sexualization) and/or report (due to different sensibilities or awareness) more harassment. These conjectures about ethnicity, foreign birth, and religion led to Hypotheses 3b and 3c (Table 1).

Method

Participants

Data were from the 1992 U.S. National Health and Social Life Survey (NHSLS). The sample was nationally representative of the adult population of the United States aged 18–59. Among the sampled individuals, 3,432 completed the interview, yielding a final response rate of 78.6%. After limiting the sample to participants ever having worked, 2,999 responses about workplace sexual harassment over the lifetime were available for analysis (1,692 for women and 1,307 for men). The interview included both initial face-to-face responses to an interviewer and later portions allowing private response to sensitive questions (in which participants read and answered the questions themselves). This article draws on the public use data set located at http://www.src.uchicago.edu/prc/nhsls.php

Procedure

Most interviewers were experienced personnel given further training in conducting interviews by the National Opinion Research Center (NORC) in Chicago and remained with the project throughout the interview period. Participant consent was obtained prior to the interview. The institutional review board of the Division of the Social Sciences at the University of Chicago approved the interview methods.

Measures

Sexual Harassment

Near the end of the interview, in the self-administered part described above, and after many other potentially sensitive questions, participants were asked about workplace sexual harassment. The wording of the question for men was as follows: “Sometimes at work, men find themselves the object of sexual advances, propositions, or unwanted sexual discussions from co-workers or supervisors. The advances sometimes involve physical contact and sometimes just involve sexual conversations. Has this ever happened to you?” The same question was asked for women. It should be noted that the inclusive nature of this question made it difficult to distinguish gender harassment from unwanted sexual attention in the analysis, the question potentially overlapping with both dimensions of harassment (Waldo, Berdahl, & Fitzgerald, 1998).

Among the independent variables described below, the following had missing values replaced with 0’s (number of substitutions in parentheses), leading to possible attenuation of effects: not U.S. born (2), education (20), and works nights/weekends (46). Additionally, only with age at first job (13), regression-based imputation was used.

Life Course Variables

The items (all based on self-report) in this group included a dummy variable indexing early puberty (≤age 11), a dummy variable for pre-pubertal sexual contact (of any kind), sex partners lifetime, with six levels ranging from none (0) to five or more (5), and education, an ordinal variable with five levels, ranging from less than high school (0) to Masters degree or higher (4).

Current Behaviors/Traits

The variables in this group were conceived as proximal indicators of sexualization. They included sexual interests and a dummy variable for frequent masturbation (≥few/month) (corresponding roughly to the 75th percentile or higher for men and 90th percentile or above for women). Sexual interests were an indicator (ranging from 0 to 13) for the total number of sexual practices (out of 14) that a participant found potentially “somewhat or very appealing.” These practices included more common ones like vaginal intercourse and watching the partner undress, as well as items like passive or active (for men) anal intercourse, having sex with multiple persons simultaneously, and forcing someone or being forced to have sex.

Occupation and Work Situation

These indicators indexed both victim-vulnerability and power-threat under the power differentials model, and a perpetrator’s opportunity for and cost of harassment, under the routine activities model. The first variables in this group indexed a participant’s usual occupation. Based on the conjectures about exposure made above, a different reference category was used for each gender. For women, the included categories were dummy variables for manual, sales and service, clerical, and upper white collar (combining administrative or managerial and professional/technical jobs), with farmer as the reference. For men, manual and farmer were combined into the reference group, with the other included variables remaining the same. The next indicator in this group was a dummy variable for currently not working, followed by works nights or weekends, an ordinal variable ranging from 0 (never) to 5 (more than twice a week). The last indicator was served in military, indicating any service, past or present, in the armed forces.

Background Conditions

The first variables in this set of controls for background conditions were for a participant’s current age. Based on prior exploratory analysis, a different reference group was used for each gender. Thus, for women, the included categories were ages 18–20, 20–29, 40–49, and 50–60, with age 30–39 as the reference. For men, 30–39 was included and 20–29 served as the reference, the other categories remaining the same. Next, (self-reported) age at first job was included as a control variable for cumulative exposure. Ethnicity was indexed as black, Hispanic, or Asian or Native American, with non-Hispanic White as the reference category. This was followed by a dummy variable for not U.S. born, indicating the participant was born outside the U.S. The final items in this group indexed a participant’s religious affiliation. Dummy variables for moderate Protestants (including, among others, Methodists, Lutherans, Presbyterians, Episcopalians, and United Church of Christ), none (i.e. no religious affiliation), Catholic, and Jewish/Other were included, with fundamentalist Protestants (including Baptists, Pentecostals, Churches of Christ, and Assemblies of God) as the reference group.

Statistical Analyses

Results were weighted in the analyses using svy methods in the STATA 9.0 statistical package, first using population weights that adjusted for the intentional oversampling of Blacks and Hispanics. Standard errors were adjusted for clustering (sampling individuals within each of 84 primary sampling units).

Results

Prevalence

Among those aged 18–60, 41% of women (CI, 37–44) and 32% of men (CI, 29–35) reported any workplace sexual harassment over the lifetime (Table 2), with the lack of overlap between the confidence intervals indicating a statistically significant gender difference in harassment. Comparing gender prevalence by age group, however, revealed a more complex pattern. Specifically, divergence in harassment prevalence only started with individuals in their 30s, with women aged 30–39 reporting significantly more harassment experiences (44%; CI, 39–50) than their male age-peers (34%; CI, 29–39). In contrast, among those 20–29, 39% of women (CI, 35–44) and 36% of men (CI, 31–42) reported having been harassed, with the overlapping confidence intervals indicating no statistically significant gender difference in harassment. The same pattern was found among the youngest participants (18–20)—although, given small cell sizes in this age group, estimates were less precise, as indicated by the large confidence intervals.

Table 2 Prevalence of lifetime workplace sexual harassment by gender and age groupa
Full size table

The results described below are for Table 3, containing survey-weighted univariate (Columns 1, 5) and multivariate (2–4, 6–8) logistic equations for predictors of sexual harassment among both women and men.

Table 3 Survey-weighted means and logistic models for lifetime workplace sexual harassment in sample aged 18–60 years
Full size table

Age-Adjusted Univariate Logistic Models

When logistic models controlling only for age were run for the 15 conditions one-at-a-time (Table 3), significance was achieved at the p < .05 level for all but four conditions for both women (column 1) and men (column 5).

Multivariate Logistic Models

Many of the same items were also significant in the multivariate results that considered the net effect of conditions taken together (Table 3). There were two sets of multivariate results: for women (Models 2–4) and for men (Models 6–8). In each set, the first results (Models 2, 6) were the simplest, containing only the early developmental variables of early puberty and pre-pubertal sexual contact, and all background conditions excepting ethnicity. The second set of results (Models 3, 7) added number of sex partners, education, and sexual interests, with ethnicity replacing religion and foreign birth among the background conditions. The final set of results (Models 4, 8) added frequent masturbation and the occupation and work situation variables.

Life Course Variables

This block of variables indexed biological transitions and/or eroticizing social experiences in adolescence (with the argument that transitioning early into puberty, or being sexually victimized before reaching puberty, may lead to a sexualized personality structure and to more conscious or unconscious sexual signaling, and thus to further victimization through harassment in later years), as well as a participant’s education, and his/her total number of spouses or cohabiting partners over the lifetime (proxying a sexualized life trajectory).

Among women, pre-pubertal sexual contact significantly elevated likelihood of harassment (OR = 2.78, Model 2), while early puberty had no significant correlation. In Model 3, education (OR = 1.19) and total number of sex partners over the lifetime (OR = 1.36) were also correlated with women’s harassment reports. Among men, in Model 6 early puberty (OR = 2.14) and pre-pubertal sexual contact (OR = 1.40), and in Model 7 education (OR = 1.24), all had the expected relationships.

Current Behaviors/Traits

This set of variables included proximal indicators of current sexualization. Among women, sexual interests (OR = 1.26, Model 3) and frequent masturbation (OR = 2.16, Model 4) were both correlated with elevated harassment reports. Similarly among men, sexual interests (OR = 1.06, Model 7) and frequent masturbation (OR = 1.54, Model 8) both had the expected relationships.

Occupation and Work Situation

As indicated above, this group of variables indexed both the vulnerable-victim and power-threat mechanisms under the power differentials approach, as well as a perpetrator’s opportunity for and cost of harassment, under the routine activities model. As per Model 4, women usually in manual (OR = 3.26), clerical (OR = 3.78), and upper white collar (OR = 3.38) jobs all reported more harassment than farmers, with working nights and weekends (OR = 1.12) also having a positive correlation. In contrast, women who were not currently in the labor force (OR = 0.73) reported less harassment. Similarly, among men in Model 8, sales and service (OR = 1.89), clerical (OR = 2.22), and upper white collar (OR = 1.49) jobs all elevated harassment reports, as did working nights and weekends (OR = 1.10) and military service (OR = 1.48).

Background Conditions

Among this set of controls for background conditions, the first was a participant’s current age. To recall, it was argued that controlling for age at first job, greater current age represented more cumulative exposure to harassment. Partially consistent with this argument, in the results (Model 2), women currently aged 18–20 (OR = 0.31) and 20–29 (OR = 0.79) reported significantly less lifetime workplace harassment than those aged 30–39, but contrary to the cumulative exposure argument, the correlations among older groups were not significantly higher. In fact, going solely by the direction of the correlations, the age pattern for women seemed curvilinear, peaking at the reference point—reflecting the prevalence patterns in Table 2. Next, in Model 3, black women (OR = 0.65) reported less harassment than white women, with foreign birth (OR = 0.45) also having a similar negative correlation in Model 2. Among the indicators of religious affiliation, as per Model 2, women with no current religious affiliations (OR = 2.77), as well as those affiliated to less conservative denominations such as moderate Protestants (OR = 1.50, Model 2) and Jewish women or those in other religious groups (OR = 2.00), all reported more harassment than women in fundamentalist Protestant denominations.

Among men, as with women, the age pattern seemed curvilinear, with those aged 50–60 (OR = 0.59, Model 6) reporting significantly less harassment incidents than men aged 20–29. Higher age at first job (OR = 0.93) had a similar correlation in Model 7. Net of other variables (Model 7), none of the ethnic categories were correlated with men’s harassment reports. However, foreign-born men (OR = 0.50, Model 6) reported less harassment. Among the religion categories, as per Model 6, men with no current religious affiliations (OR = 1.51) and Catholic men (OR = 1.51) both reported more harassment than their fundamentalist Protestants counterparts.

Supplementary Analysis: Homosexuality

Due to high collinearity with the predictors in Table 3, the variables in this set were left out of the multivariate analysis, and added one at a time to models including only linear and quadratic controls for age. They included homosexual self-identification, any lifetime homosexual activity (both dummy variables), and current appeal of same-gender sex, an ordinal indicator ranging from 1 (not at all appealing) to 4 (very appealing). For women, lesbian self-identification did not have any correlation with harassment. However, both lifetime homosexual activity (OR = 2.08, p < .05) and current homosexual appeal (OR = 1.36, p < .01) were strongly correlated with women’s harassment. Among men, in contrast, while gay self-identification (OR = 2.27, p < .10) was mildly correlated with elevated harassment reports, neither lifetime homosexual activity nor current appeal had a significant effect.

Supplementary Analysis: Pre-Pubertal Sexual Contact

Due to results substantively similar to those for any pre-pubertal sex, and smaller cell sizes, these variables were not included in the multivariate analysis. As with the homosexuality variables, the two indicators in this set were added one at a time to models including only linear and quadratic controls for age. They included a dummy variable for any pre-pubertal anal sex, and an ordinal scale denoting number of sexual partners before puberty—ranging from 0 (none) to 3 (3 or more). For women, number of pre-pubertal partners had a strong correlation with lifetime harassment (OR = 1.89, p < .01), while among men, both anal sex (OR = 3.58, p < .01) and number of partners (OR = 1.26, p < .05) were correlated with elevated harassment reports.

Discussion

Prevalence

As noted, the non-overlapping confidence intervals for harassment prevalence among women and men aged 18–60 (Table 2) indicated a statistically significant gender difference. To quantify the magnitude of this gender effect, a separate logistic model (not shown) was run, regressing sexual harassment on a dummy variable indicating whether the participant was a woman (1) or a man (0). As per the results (OR = 1.46, p < 0.01), women were 46% more likely than men to report any lifetime workplace harassment. However, as suggested by the overlapping confidence intervals for women and men in their 20s or younger, harassment prevalences for the two genders may have converged over time, with men in these more recent cohorts no less likely than their female age-peers to have experienced workplace harassment.

Models and Mechanisms

This study began with two contrasting theoretical models of harassment, the power differentials approach—emphasizing asymmetries in formal or informal power—and a routine activities model borrowed from criminology, emphasizing risk factors emerging from a victim’s attributes or behavior and from the situational context. Overall, the harassment results (Table 3) were more uniformly consistent with a routine activities pattern.

Previous studies based on the power-differentials model have conceptualized women’s harassment as an exercise in dominance by individuals, usually men, in positions of greater power. Hypothesis 1a (Table 1) reflected this “vulnerable victim” proposition. As per the second, “power threat” mechanism (Hypothesis 1b) under the power differentials model, assertive women, such as those in high status jobs, would be denigrated through harassment. In the results, among women, all of the included occupation categories (manual, sales and service, clerical, and white collar) had roughly equivalent correlations with harassment reports, relative to women in farming jobs. In other words, neither the vulnerable victim nor the power threat mechanism seemed to explain women’s harassment experiences, with neither women in positions of lesser power in the occupational hierarchy, such as manual labor (vulnerable victim), nor those in white-collar occupations (power threat), particularly distinctive in reporting more harassment. Next, it was argued that abstracted away from its traditional focus on women’s harassment, the vulnerable victim mechanism could potentially apply to the harassment of men in less powerful positions as well, whether by powerful women or men. As with women, however, all the included categories of “office jobs” (sales and service, clerical, and white collar) had roughly the same correlation with men’s harassment reports, relative to manual workers and farmers, suggesting perhaps that victim-vulnerability and a perpetrator’s greater power may not play a major role in men’s harassment.

However, in the supplementary age-adjusted homosexuality analysis presented above, gay men were more likely to report harassment, with this correlation reaching significance despite a small cell size. This last result suggests that hostile sexism—perpetrated by dominant heterosexual men upon women as well as weaker, more effeminate or homosexual males as a means of defending traditional gender roles—may be one mechanism driving men’s harassment. Inclusion of this variable in the multivariate models for men did not lead to alterations in the magnitude of other correlations. Additionally, in separate cross-tabulations, only 3% of harassed men of all ages identified themselves as gay, 7% reported any lifetime homosexual activity, and 5% found same-gender sex potentially “somewhat” or “very appealing.” Intriguingly, the homosexuality results did suggest a strong effect of hostile sexism on women’s harassment, with women reporting any lifetime same-gender sexual experiences or current appeal markedly more likely to have been harassed. As with men, however, only small proportions of harassed women were self-identified lesbians (2%), reported any lifetime homosexual activity (5%), or current appeal (7%). Nor did the inclusion of either of these variables in the full models for women affect the magnitude of other correlations. Hence, while acknowledging the possible relevance of hostile sexism, it should be noted that this mechanism failed to account for most women’s or men’s harassment experiences.

As noted, the results were more uniformly consistent with the routine activities model, integrating a perpetrators perceived benefit from and opportunity for harassment, as well as lower cost in the form of third-party sanctions. With regard to the first mechanism (benefit), it was argued that both women and men with more sexualized personality structures may send out, consciously or unconsciously, cues that are perceived as indicating receptivity to sexual advances. It was noted that the NHSLS data did not contain any direct indicators of such sexual signaling by a target, but that sexualization itself, as a precursor to signaling, could perhaps derive partially from early developmental factors, such as childhood sexual contact and early puberty (Hypothesis 2a) and from more education, and that patterns such as multiple sex partners over the lifetime, diversified sexual interests, and frequent masturbation might be good indicators of such a sexualized dispositional structure (Hypothesis 2b). Consistent with these conjectures, five out of these six variables (excluding early puberty) were strongly correlated with elevated harassment reports among women, and five (excluding sex partners) among men. Also, when an additional sexualization indicator, “ever giving a man oral sex”—left out of the main analysis due to high collinearity with other sexualization variables—was added to the first set of multivariate models (2, 6), the correlation for women (OR = 2.88) but not men was significant (at p < .01). The linkage with any pre-pubertal sexual contact, in particular, suggests a different—and perhaps richer—conception of victim vulnerability to harassment among both women and men, one based not simply on current power asymmetries, but on cumulative victimization over the life course. This last conclusion was also supported by the supplementary analysis presented above, which found strong correlations of harassment with pre-pubertal anal sex (men only) and more partners before puberty (both genders). However, since none of these indicators contained direct information on coercion, the “victimization” inference remains tentative. Additionally, in separate analysis (not shown), the results for lifetime sex partners remained robust when this ordinal variable was replaced with dummy variables indicating none, one, and two or more partners, suggesting that these correlations were not driven by a few outliers with many partners. Education, of course, might also sensitize an individual to the harassment issue, increasing one’s propensity to label the same behaviors as harassing. It was not possible to tease apart these different mechanisms in the analysis.

With regard to a perpetrator’s increased opportunity for harassment, the second mechanism under the routine activities model, it was argued that women usually in non-farming occupations and men in office jobs, as well as men with any military service, would face greater exposure to potential harassers and, therefore, report more lifetime harassment (Hypothesis 2c). The results were consistent with this hypothesis. Among women, those in all non-farming occupations other than sales and service reported more harassment than women in farm jobs, with many of the latter presumably working on family farms and thus facing the least exposure to strangers. Similarly, all men in office jobs (more exposed to all harassers than farmers, and to opposite-gender harassers than manual workers) as well as those with military service (living and working in close proximity with others) reported significantly more harassment. It is possible, however, that the manual labor or farming category among men simply indexed less sensitivity to sexual attention and a consequent underreporting of such incidents as harassment. Ethnographic literature on blue-collar occupations suggests, for instance, that sexual banter is a familiar part of workplace interaction in such settings (Carey, 1994; Walshok, 1987; Yount, 1991).

The third mechanism under the routine activities model was a perpetrator’s lower cost of harassment. It was argued that individuals routinely working nights or weekends might be exposed to more situations where a potential harasser is free from observation and sanctions from third parties. Accordingly, Hypothesis 2d (Table 1) was included. The results were consistent with this hypothesis among both genders.

To summarize, gross differences in power, as indexed by broad occupational categories, were not correlated with workplace harassment among American women and men, with neither women nor men at the bottom nor women at the top of the occupational hierarchy particularly distinctive in reporting more harassment. Contrary to this power differentials approach, still dominant in the harassment literature, the occupational results suggested instead that harassment—among both American women and men—may be driven at least as much by such “routine activities” mechanisms as more opportunity and lower cost for perpetrators. At least with regard to women’s harassment, it should also be noted that previous studies asserting a power differentials pattern have often been speculative or synthetic essays (Cleveland & Kerst, 1993; Robbins, Bender, & Finnis, 1997; Wilson & Thompson, 2001), while others have been based on small local samples such as nurses in a hospital (Cholewinski & Burge, 1990; Finnis & Robbins, 1994; Grieco, 1987) or samples of special populations such as military personnel (Harned, Ormerod, Palmieri, Collinsworth, & Reed, 2002; Martindale, 1991), as opposed to nationally representative data such as the NHSLS. A notable exception is Uggen and Blackstone (2004), based partly on the 1996 round of the U.S. General Social Survey (GSS), which found lack of power to be positively associated with workplace sexual harassment in a combined analysis of women and men. However, this brief GSS analysis operationalized lack of power through subjective (dis)satisfaction with one’s current financial situation–arguably a less direct proxy than the objective occupational status indicators in the present study.

Finally, with regard to background conditions, it was argued that controlling for age at first job, greater current age could index more cumulative exposure to harassment. In other words, it was expected that individuals in the oldest cohort (50–60) would report the most lifetime workplace harassment. Additionally, it was argued that, especially among women, being currently out of work could be a marker of episodic entry into the labor force (and hence shorter time at risk), and thus be correlated with less reports of lifetime harassment. In other words, both the age and employment variables also indexed opportunity (under the routine activities model) in a temporal sense. Hypothesis 3a reflected these propositions. The results for age ran partially counter to these expectations, especially among men, with men in their 50s—despite having faced more years at risk—significantly less likely to report harassment experiences than men who had entered in their 20s and joined the work force in the late 1980s and early 1990s. Although not controlling age at first job, the prevalence-patterns in Table 2 also supported this observation. This last result seemed indicative of cohort effects, consistent with the conjecture, supported by media reports (Armour, 2004; Gross, 1995; Lawlor, 1994), that male workplace harassment is a relatively recent phenomenon. It is also possible, however, that norms of masculinity internalized by older cohorts hindered them from reporting—or perhaps even labeling—unwanted sexual attention as harassing or that such incidents were not as traumatic as for women and therefore more easily forgotten with time (Waldo et al., 1998). Consistent with the argument about shorter time at risk, however, women currently out of the labor force reported significantly less harassment.

Next, it was speculated that black women might report less harassment, and that foreign birth could imply less sensitivity to sexual attention, i.e., that individuals from societies with more liberal norms about workplace sexual attention would be less likely to label such behaviors as unwanted or offensive (Hypothesis 3b). In the results, black women were less likely to say they were harassed than their non-Hispanic white counterparts, although whether this was due to underreporting out of low expectations of community support remains unclear. Recent research using focus groups to analyze the linkage between race and harassment definitions (Welsh, Carr, MacQuarrie, & Huntley, 2006) suggests that the mechanism may instead be a “normalization” of sexual attention among both black women and men, a notion supported by the negative direction of the effect among men. Finally, both foreign-born women and men reported significantly less harassment, consistent with the conjecture about lower sensitivity.

Finally, with regard to religion, it was speculated that those with no current religious affiliations and those affiliated with non-conservative religious groups would perhaps experience and/or report more harassment (Hypothesis 3c). The patterns for both women and men were largely consistent with this conjecture, with both women and men with no religious affiliations, moderate Christian women, and Jewish women as well as women belonging to non-mainstream religious groups markedly more likely to say they were harassed than those with fundamentalist Protestant affiliations. However, Catholic men also reported more harassment incidents—an unanticipated result that is presented here as a purely inductive finding. It should also be noted that the correlations for moderate or no religious beliefs could index either more harassment experiences (through the sexual signaling mechanism), or greater sensitivity to the harassment issue (due to different sensibilities in general), with these separate mechanisms difficult to tease apart in empirical analysis.

Limitations

There were several limitations to these analyses. Most importantly, given the cross-sectional data and the nature of the dependent variable (workplace harassment experiences over the lifetime), temporal order and causal direction could not be demonstrated. Additionally, the NHSLS data contained no “local” information, either about a victim’s greater exposure to potential harassers (increased opportunity for a perpetrator) or his/her power or status relative to people she/he came in regular contact with at the work-site (vulnerable victim, power threat). Hence, inferences about these three mechanisms were based on a victim’s own broad occupational position. Similarly, given a lack of direct indicators for a victim’s sexual signaling, indirect proxies for his/her sexualization had to be used. The NHSLS indicators for pre-pubertal sexual contact contained no direct information on coercion. Hence, conclusions about the “victimization” trajectories remain tentative. Next, the inclusive nature of the harassment question made it impossible to distinguish gender harassment from unwanted sexual attention, the question potentially overlapping with both dimensions of harassment. Hence, possible distinctions in the qualitative nature of harassment experiences for the two genders could not be analyzed. The analysis was based on self-reports, which provided no direct evidence of harassment patterns, making participants’ differential sensitivity to the same behaviors a potential problem. In particular, more sexualized men and women, rather than facing more harassing incidents, may simply have a greater propensity to frame interactional cues as sexual—and, when unwelcome, as harassment. Cell sizes were small for some of the analyses (e.g., childhood sexual contact, and military service among women), making it difficult to reliably distinguish signal from noise. Finally, the NHSLS data are now 15 years old. It is acknowledged that workplace harassment patterns may have evolved since then–especially among men, for whom the results suggested cohort effects.

Summary

Data from a nationally representative sample of the United States suggested that while, among those aged 18–60, workplace sexual harassment over the lifetime was significantly more prevalent among women than men, the two genders’ harassment prevalences may have converged in more recent cohorts. Specifically, men’s harassment appeared to be a more recent phenomenon, with men in younger cohorts reporting more lifetime workplace harassment than older men, despite the latter having faced more years at risk. Moreover, the structure of risk factors underlying harassment was similar for women and men. Contrary to much of the harassment literature, among both genders workplace harassment seemed to have at least as much to do with a system of “routine activities” mechanisms—a victim’s conscious or unconscious sexual signaling and more exposure to harassers, and a perpetrator’s lower cost of harassment—as with (unobserved) victim–perpetrator differentials in power. Intriguingly, the strongest risk factors among both genders were related to sexualization, with the linkage to pre-pubertal sexual contact in particular suggesting, for both women and men, a richer conception of victim vulnerability—one based not simply on current power asymmetries but on cumulative victimization over the life course. Reports of harassment experiences were also linked to a victim’s higher exposure to potential harassers and a perpetrator’s lower potential costs in the form of third party sanctions, with women in non-farming occupations and men in office jobs, men with some military service, and both women and men working nights and weekends, more likely to report harassment. Finally, black women and individuals born in non-U.S. societies were less likely to report harassment, while the inverse was true for those with either no religious affiliations or connections with more moderate denominations.