In recent years, the academic discussion about gender equality in the labour market has come to focus on work conditions (Angelici and Profeta 2020). A theoretical starting point for this discussion is that men and women prefer work conditions of different types. Women, who have more responsibilities for children and the household, prefer jobs with more flexibility and shorter commutes. To get these good work conditions (so-called work amenities), they accept jobs with lower wages (Goldin 2014, Wiswall and Zafar 2018, Mas and Pallais 2017, Carta and De Philippis 2018, Le Barbachon et al. 2019).
We argue that work conditions may also cause gender inequality via an inefficient, discriminatory process. We look at sexual harassment as a work condition to show how this discrimination works (Folke et al. 2020). In nationally representative data from Sweden, we study which men and women face the largest risk of sexual harassment from colleagues and managers. Then, we use a survey experiment to measure how men and women valuate these risks and finally, we compare wages and job satisfaction to test if employers provide compensation to at-risk workers.
Systematically higher risk for gender minorities
To describe harassment risks, we use survey data on self-reported sexual harassment from the Swedish government’s biannual survey on work conditions (N=50,000). This nationally representative survey contains questions on unwanted sexual advances, sexist hostility, and gender harassment from colleagues or managers in the last 12 months. We use administrative data for the full Swedish workforce to compute the share of men in each survey respondent’s occupation and workplace.
The relationship between self-reported harassment and sex ratios are shown in Figure 1. Clearly, both women and men self-report more harassment when they are the gender-minority in their occupation or workplace.
Figure 1 Sexual harassment across occupations and workplaces
Notes: The figure shows binned averages of a binary variable for self-reports of sexual harassment from colleagues or managers in the last 12 months.
The higher self-reported rate of harassment among gender minorities is not caused by systematically different demographic traits. Nor is it caused by gender minorities being more likely to have opposite-sex supervisors, by themselves holding subordinate or supervisory positions (Folke et al. 2020), or by them having opposite-sex managers.
It is possible that gender minorities are more likely to complain untruthfully about their work conditions in the survey. This seems unlikely, however, because there are no relationships between sex ratios and other self-reported workplace problems like bullying, conflicts with colleagues, or lack of managerial support.
Valuations in a hypothetical job-choice experiment
We design a conjoint survey experiment with fictional job choices to measure how people valuate sexual-harassment risks in a workplace. Approximately 4,000 employed Swedish respondents chose between jobs with randomised levels of wages, skill development, schedule flexibility, and working environments. For the work environment, we randomised among the four options: 1) “No information about the work environment”, 2) “Workers are content with the work environment”, 3) “Some employees in the work unit have had conflicts with the manager”, and 4) sexual harassment – but not using these words.
To describe sexual harassment, we used three vignettes of specific and common incidents that had taken place in the (fictional) work unit. One vignette told of groping, another of unwanted and crude sexual conversation, and the third of the attitude that people of a specific gender were unsuitable for the job. We assigned the sex of the victim and sex of the perpetrator in the vignettes to the most realistic situation in the respondent’s current workplace. People in women-dominated workplaces saw male victims and female perpetrators, and vice versa for mixed and male-dominated workplaces.
Figure 2 Wage-cut equivalents of avoiding a job in a work unit with sexual harassment
Notes: Willingness-to-Pay Estimates from a conjoint survey experiment for hypothetical job choices. The reference category for sexual harassment in the work unit was to have “No information” about the work environment in the fictional job. (*) this estimation uses only data from vignettes where the respondent had the same sex as the fictional harassment victim.
Men and women had similarly large dislikes for workplaces with sexual harassment, both equivalent to the dislike of a ~10 ppt wage-cut (see the left-hand side of Figure 2). But the dislike for workplaces with sexual harassment depended strongly on the sex of the fictional victim. Respondents strongly disliked workplaces where the victim was the same sex as themselves but expressed less dislike when the victim was of the opposite sex (see the middle section of Figure 2).
Interestingly, workplaces with same-sex victims were not equally off-putting to all but depended on respondents’ own knowledge about cases of sexual harassment in their industry (right-hand side of Figure 2). People with knowledge of cases expressed a huge reluctance to enter a firm with a same-sex victim, while a same-sex victim had little importance for unaware respondents. Arguably, hearing other victims’ stories of sexual harassment is key to understanding the true harms of victimisation.
(No) compensating differentials
Having established that workers have a large, negative valuation of sexual harassment risks in the workplace, we turn to study if employers compensate them for this disutility. First, we use wage data from tax records to analyse pay compensation for the risk of sexual harassment (which we predicted for everyone in the workforce). This analysis of wage differences across and within firms with varying harassment risks did not produce any evidence of compensating pay. If anything, high-risk gender minorities earn less, not more, than their similarly qualified colleagues of the opposite sex.
It is possible that at-risk workers are compensated by in-kind payments or favourable work conditions rather than wages. To study this question we compare rates of job dissatisfaction, self-reported intentions to quit one’s job, and job separations among people who self-reported sexual harassment or not in the Work Environment Survey. These measures summarise people’s total compensation from a job. Arguably, if workers were compensated for their experiences of sexual harassment, they would not express more dissatisfaction or intentions to quit than others.
The results on job dissatisfaction and quit intentions contradict the notion of employer compensation. Both men and women who self-report sexual harassment have twice the rate of job dissatisfaction and quit intentions than men and women who do not. This gap remains large and statistically significant even after holding constant the sex ratio of the workplace, the occupation, the respondent’s demographic traits, and their self-reports of other types of workplace incivility like bullying and conflicts.
For job separations, women who self-report sexual harassment are more likely to leave their workplaces, and their job switch tends to move them to a more women-dominated organisation than the one they left. For men, there is no strong link between harassment and actual separations.
For non-economists, it may be laughable that employers would actively compensate workers for the existence of sexual harassment risks or for actual harassment incidents. It is a well-established fact that employers usually trivialise sexual harassment and often retaliate against victims when they come forth with accusations (reviewed by Cortina and Magley 2003).
A result from our survey experiment—that people care less about harassment of opposite-sex people when they look for jobs—may shed further light on the lack of compensation. Investments in harassment prevention, or direct compensation to victims, becomes irrational for employers with access to workers who are largely indifferent to the type of harassment that exists in the firm.
Conclusions
Economists should pay more attention to discrimination in work conditions and particularly to sexual harassment, which is closely related to gender inequality on the labour market. The harassment is concentrated against gender minorities, both male and female, in occupations and workplaces. These minorities suffer a substantial disutility, a disutility for which they are not compensated by employers. The systematic mistreatment of gender minorities will shape their trade-offs as they chose where to work.
Women face added costs from working in male-dominated contexts, which usually offer higher wages. Men face costs when they chose women-dominated occupations and workplaces, and where they help break the sex segregation of the labour market. Future research should look into discrimination in work conditions against other minorities on the labour market, for example by race/ethnicity, age, or sexual orientation.
References
Angelici, Marta, and Paola Profeta (2020), “Smart-working: Work flexibility without constraints”, VoxEU.org, 28 March.
Carta, Francesca, and Marta De Philippis (2018), “Commuting time and family labour supply decisions”, VoxEU.org, 11 November.
Cortina, Lilia M, and Vicki J Magley (2003), “Raising voice, risking retaliation: Events following interpersonal mistreatment in the workplace”, Journal of Occupational Health Psychology 8(4): 247–65.
Folke, Olle, Johanna Rickne, Seiki Tanaka and Yasuka Tateishi (2020), “Sexual harassment of women leaders”, Daedalus 149(1): 180–97.
Goldin, Claudia (2014), “A grand gender convergence: Its last chapter”, American Economic Review 104(4): 1091–119.
Le Barbachon, Thomas, Roland Rathelot and Alexandra Roulet (2019), “Gender differences in job search: Trading off commute against wage”, mimeo.
Mas, Alexandre, and Amanda Pallais (2017), “Valuing alternative work arrangements”, American Economic Review 107(12): 3722–59.
Wiswall, Matthew, and Basit Zafar (2018), “Preference for the workplace, investment in human capital, and gender”, The Quarterly Journal of Economics 133(1): 457–507.
