Gender,Sexual Orientation,and Behavioral Norms in the Labor Market

Abstract

The author examines bias and behavioral norms based on sex and sexual orientation in the labor market. Using an online laboratory setting, participants were asked to evaluate résumés that were manipulated on sex, perceived LGBT status, and use of traditionally masculine or feminine adjectives. Findings show that male participants penalized résumés that included an LGBT activity, and the penalty was slightly stronger for male résumés. Additionally, men evaluated non-LGBT women who used feminine adjectives more positively than when they used masculine adjectives. Résumés of women with the LGBT activity and men were both immune to this effect. This outcome suggests that perceived-heterosexual women are discouraged from masculine behavior that would be rewarded in the labor market, whereas perceived-LGBT women are not. Men who had the strongest reaction to perceived-heterosexual women using masculine adjectives also had the strongest negative reaction to résumés with an LGBT activity. This pattern suggests that male decision makers are biased in ways that harm LGBT men, LGBT women, and heterosexual women in the labor market.

Keywords

gay/bisexual men and women gender gender discrimination gender pay gap gender wage differential

Gay men and lesbians have outcomes in the labor market that differ from the outcomes for heterosexual men and women. In addition, the outcomes for gay men and lesbians are different from each other. For example, research on earnings differences consistently find that gay men appear to earn less than do similarly educated heterosexual men, whereas lesbians earn more than what similarly educated heterosexual women earn (Klawitter 2015). Relatedly, résumé audit studies find that gay men experience more discrimination in traditionally masculine fields or for job advertisements using masculine words, and lesbians face more discrimination in traditionally feminine fields (Tilcsik 2011; Ahmed, Andersson, and Hammarstedt 2013). The distinct experiences of gay men and lesbians in the labor market suggest a nuanced story of differences in human capital, intra-family decisions, and discrimination.

Literature from psychology reveals cognitive biases that affect how women and sexual minorities are perceived and that shed light on these differing labor market outcomes for gay men and lesbians. First, people often hold strong stereotypes about personality attributes based on sex and sexual orientation. Second, whereas both heterosexual men and women often have negative reactions to gay people, heterosexual men have stronger negative reactions to gay people than do heterosexual women and often have stronger negative reactions to gay men than to lesbians. Third, a broad literature has established that male laboratory participants generally prefer women who behave in traditionally feminine ways to those who behave in traditionally masculine ways.

These three patterns suggest the need for a deeper analysis of how gender and sexual orientation influence social interactions, particularly their implications in the labor market. In this study, I examine whether men and women evaluate job applicants unequally based on the applicant’s sex, sexual orientation, and use of masculine or feminine language. I asked participants in an online laboratory setting to evaluate résumés that vary based on sex, including a lesbian, gay, bisexual, or transgender (LGBT) extracurricular activity, and the use of traditionally masculine or feminine adjectives.

The results of the laboratory experiment show that male participants rated both men and women with an LGBT activity as less employable than applicants with a non-LGBT activity. This effect was somewhat stronger for men with the LGBT activity. Male participants prefer women who use feminine language to those who use masculine language only if those women do not have an LGBT activity on their résumé. Women with an LGBT activity on their résumés are immune to this effect. Notably, men, both with and without an LGBT activity, do not experience any difference when they use masculine language on their résumés. Female participants, by contrast, only mildly differentiate between applicants based on sex, sexual orientation, and their choice of masculine or feminine adjective.

Motivating Literature

Differences in Labor Market Outcomes Based on Sex and Sexual Orientation

Along with the well-established differences in earnings between women and men (Blau and Kahn 2017), consistent evidence shows that gay men earn less than heterosexual men earn (Badgett 1995; Carpenter 2007; Elmslie and Tebaldi 2007; Klawitter 2011; Martell 2012). Many studies also find that gay women earn substantially more than do heterosexual women (Black, Gates, Sanders, and Taylor 2000; Berg and Lien 2002; Black, Makar, Sanders, and Taylor 2003; Blandford 2003; Jepsen 2007; Antecol, Jong, and Steinberger 2008; Cushing-Daniels and Yeung 2009). There are notable exceptions: Badgett (1995) and Carpenter (2008) found a lesbian penalty, Sabia (2014) found no lesbian effect, and Carpenter (2005) and Frank (2006) found no difference for gay men. A recent meta-analysis of 31 studies on earnings differences found an average earnings premium of 9% for lesbians and an average earnings penalty of 11% for gay men (Klawitter 2015). As shown in Online Appendix A, these patterns remain consistent using 2016 data.

Men in same-sex couples are disproportionately represented in female-dominated occupations, and women in same-sex couples are more likely to work in male-dominated occupations. This outcome is consistent with findings from audit studies that gay men face more discrimination in traditionally male fields, and gay women face more discrimination in traditionally female fields (Ahmed at al. 2013). As shown in Figure 1, the estimated proportion of men who are in a same-sex couple in each occupation increases dramatically among occupations that are heavily female. The reverse pattern holds for women, with the estimated proportion of women who are in a same-sex couple in each occupation being highest among occupations that are heavily male.

Figure 1.

Occupations for Individuals in a Same-Sex Couple

The observed differences in income, labor force participation, and occupations for gay men and lesbians may be driven by differences in human capital. For example, lesbian and bisexual women may anticipate not having a higher-earning partner, specifically a higher-earning male partner, and therefore invest in education in fields with higher-earning potential, which are typically male-dominated fields (Klawitter 2015). A second commonly examined theory is the division of household labor—same-sex couples tend to have more egalitarian divisions of labor (Carrington 1999). Less-explored explanations for these patterns are social norms and cognitive biases that affect how LGBT men and women are perceived by employers.

Insight from Psychology on Labor Market Differences

Psychology and behavioral economics offer important insight into labor market outcomes based on sex and sexual orientation. First, heterosexual men and women often have negative reactions to gay people, but heterosexual men often have stronger negative reactions to gay people than do heterosexual women. Also, heterosexual men often have stronger negative reactions to gay men than to lesbians (Kite and Whitley 1996; Raja and Stokes 1998; Herek 2000, 2002; Gough 2002; Moskowitz, Rieger, and Roloff 2010).

People often hold stereotypes about personality characteristics based on sex and sexual orientation. For example, women are perceived as being more cooperative, sensitive, and affectionate whereas men are seen as more independent and assertive (Broverman et al. 1972; Deaux and Lewis 1984; Heilman 2001; Heilman and Parks-Stamm 2007). These descriptive stereotypes may cause people to anticipate a “lack of fit” between a heterosexual female applicant and a job that is perceived to require masculine traits (Heilman 1995; Weichselbaumer 2003, 2004).

People often believe that gay men and lesbians hold more personality attributes typically associated with the opposite sex, for example, that lesbians have more masculine traits than do heterosexual women, and gay men more feminine traits than do heterosexual men (Kite and Deaux 1987; Ahmed et al. 2013). Likewise, men with feminine traits and women with masculine traits are perceived as more likely to be gay (Deaux and Lewis 1984). This tendency appears to affect labor market interactions, wherein employers in male-dominated fields or that include masculine traits in the job posting are less likely to contact a gay male applicant, and employers in female-dominated fields are less likely to contact a lesbian applicant (Tilcsik 2011; Ahmed et al. 2013).

Historically, a wide range of gender expression exists within the LGBT community (Halberstam 1998). LGBT slang is a method to name culturally specific ideas, and it also serves as a tool to subtly identify other members of the LGBT community (Proschan 1997). Terms such as “femme,”“lipstick lesbian,”“chapstick lesbian,” and “butch” refer to variations in physical expression of gender among queer women. Among queer men, terms such as “butch,”“bear,”“otter,”“swish,” and “twink” differentiate between gay men of different physical appearance. Some research on LGBT discrimination has included photos and other signals of physical appearance along the butch–femme dichotomy (Weichselbaumer 2004); however, in this article I focus exclusively on personality traits rather than physical appearance.

Prescriptive stereotypes—stereotypes about how a woman ought to be—often cause employers or coworkers to react negatively when women violate these stereotypes (Heilman 2001; Rudman and Phelan 2008). For example, a laboratory study found that being described as a “successful manager” increased the perceived competence and independence of both men and women (Heilman, Block, and Martell 1995). Yet, this also negatively affected women: Respondents reported that women are less “hostile to others” than are men in general, but women who are “successful managers” are viewed as more “hostile to others” than are men who are described as “successful managers” (Heilman et al. 1995). More broadly, respondents often react negatively when women engage in traditionally masculine actions in the workplace, including withdrawing altruistic behavior, being successful in a male occupation, and self-promotion in an interview (Rudman 1998; Rudman and Glick 1999, 2001; Gill 2004; Heilman, Wallen, Fuchs, and Tamkins 2004; Heilman and Chen 2005). This negative reaction has been found across a variety of experimental manipulations and among different groups of respondents.

One example of this pattern is how participants react to women who attempt to negotiate for a higher salary (Bowles, Babcock, and Lai 2007; Amanatullah and Morris 2010). Both male and female laboratory participants were less likely to want to work with women who negotiated, and participants also described the women as less nice and more demanding, although equally competent. While men were also viewed as less nice and more demanding when they negotiated, there was no corresponding change in male participants’ willingness to work with them. This result suggests that women are penalized for negotiating because negotiating violates a prescription of femininity: niceness (Bowles et al. 2007).

In a related literature on prosocial behavior, laboratory participants generally help women more than they help men; this effect is particularly strong for male participants, who may want to be protective, heroic, or chivalrous (for a review see Eagly and Crowley 1986 and Eagly 2009). Rather than punishing women who act in a counter-stereotypical way, participants may want to act in a chivalrous or protective way toward women who act in a feminine way. Indeed, Bowles et al. (2007) found that women who did not negotiate were rated more highly than were men who negotiated as well as men who did not negotiate. Although this suggests that the “backlash” could be interpreted as a form of chivalry, the result is the same: discouraging women from behavior that is rewarded in the workplace.

In this article, I consider three closely related questions. First, I examine if participants show bias toward perceived LGBT résumés, and if more bias is shown toward male LGBT résumés than to female LGBT résumés. That is, I test if the documented higher rates of homophobia toward men also result in more negative assessment of their résumés. Second, I examine if the same behavioral prescriptions that face heterosexual women apply equally to LGBT women. Third, I test if the first two questions are driven by male or female participants. To assess these three inter-related questions, I test how people perceive résumés that vary on sex, LGBT status, and the use of traditionally masculine or feminine language.

Methods

Experimental Manipulation

I created 10 résumés based on a compilation of résumés from recent college graduates who publicly listed their résumé on Indeed.com, similar to the compilation of résumés used by Bertrand and Mullainathan (2004). Indeed.com is an online job posting site, similar to Monster.com and CareerBuilder.com. Applicants can post résumés for free on Indeed.com, which are publicly accessible. The 10 résumés were created from randomly selected résumés of people with a recently awarded bachelor’s degree in biology from among those listed on Indeed.com on a specific date (October 30, 2013) in Durham, North Carolina. Unlike most fields, bachelor’s degree recipients in biology are neither over- nor under-represented among women; women earned 57.2% of all bachelor’s degrees in 2010 and 57.8% of all bachelor’s degrees in biological and agricultural sciences (National Center for Science and Engineering Statistics 2013).

Each compilation résumé contains randomly selected elements of each randomly selected résumé. That is, a résumé contains the university name from one résumé, job title and description from another, another job from a third, and so forth. An annotated example résumé is included in Online Appendix B.

The objective statement of the résumé, a common feature on résumés of recent college graduates, includes a mix of adjectives that are either masculine or feminine. Adjectives were selected from a pre-test that determined which adjectives are perceived as masculine—in this context “masculine” is defined as being traditionally associated with men. In the pre-test, one group of participants on Mechanical Turk (or MTurk, described in more detail below) viewed adjectives that were supposedly from a résumé and answered the question “How likely is it that the applicant is male?” Another group rated the same adjectives on how likely the applicant was female. As Figure 2 shows, adjectives that were viewed as relatively more likely to come from a male applicant by one group were viewed as less likely to come from a female applicant by the other group. This suggests that the manipulation will be effective; that is, using adjectives perceived as the most feminine and least masculine will signal traditionally feminine characteristics. Likewise, using adjectives perceived as the most masculine and least feminine signals traditionally masculine characteristics. For this study, I consider masculine adjectives to be aggressive, enterprising, assertive, bold, confident, self-starter, achiever, and dynamic; and feminine adjectives to be nurturing, caring, sympathetic, kind, supportive, encouraging, helpful, and cooperative.

Figure 2.

Results from a Pre-Test of Adjectives

As shown in Online Appendix D, 5.4% of publicly listed résumés located in Durham, North Carolina, in May 2018, with a degree in biology contain the words from the feminine manipulation. Moreover, this percentage is consistent for all experience levels, indicating that the use of feminine adjectives is neither uncommon nor evidence of inexperience in this labor market.

On the résumé, the applicant’s sex is indicated by the applicant’s first name. The choice of a name is complicated by the fact that names also imply information about race. How favorably participants view résumés of gay applicants differs if the résumé appears to be from an African American applicant rather than from a white applicant (Pedulla 2014). Pedulla (2014) found that participants rated the résumés of white gay men less favorably than they rated résumés of white heterosexual men, but the reverse pattern held for African American male résumés. For this project, I restricted myself to using names that are more common among white people and will leave variation in the effect by race to future work. The first names used in the manipulations are common among babies born to white, highly educated parents in the 1990s (so the individuals would have been in their early twenties during the time of the study). The female names are Katherine, Emma, Alexandra, Julia, and Rachel (Levitt and Dubner 2005). The male names are Benjamin, Samuel, Alexander, John, and William (Levitt and Dubner 2005). Out of the 100 most common last names from the 2000 Census, I selected the last names with the highest percentage of white people. These last names are Wood, Sullivan, Myers, Peterson, Miller, Murphy, Fisher, Cox, Cook, and Long (Census Bureau 2012).

I manipulated the way sexual orientation would be perceived in each résumé by including a leadership position in a college group. Some résumés indicated the applicant held a leadership position in an LGBT group, whereas others indicated the applicant held a similar role in a non-LGBT organization. For example, one non-LGBT activity was labeled “Student Activities Board” and described how the applicant “planned and organized events promoting diversity.”Tilcsik (2011) performed an audit study comparing callback rates for résumés of men that indicated they were the treasurer of a campus LGBT organization to those that indicated being the treasurer of a campus socialist organization. He found that 11.5% of the résumés with the socialist organization received a callback compared to 7.2% for the résumés with the LGBT organization. This outcome suggests that listing membership in a college LGBT organization on a résumé is noticed by potential employers.

I recruited participants on Amazon Mechanical Turk (MTurk) to assess the résumés on personality characteristics and level of perceived skill. MTurk is a marketplace that pays piece rate for small tasks completed online. Some studies have shown that participants from these samples are not representative of the entire US population in terms of age (MTurk is skewed toward younger participants), but they are a closer representative distribution of race in the US population than are the study participants typically recruited on college campuses, and their responses are reliable (Buhrmester, Kwang, and Gosling 2011; Berinsky, Huber, and Lenz 2012; Horton, Rand, and Zeckhauser 2011). I restricted participants to those with an IP address in the United States and who had already successfully completed a specified number of tasks for other employers on MTurk.

The participants were told they were helping a company sort résumés for an entry-level position for a college graduate who majored in biology. Concealing that the task was part of a research study reduced the chance that participants would alter their behavior to avoid appearing discriminatory or to “help” the researcher obtain the desired results. This concern is especially pertinent for workers on Mechanical Turk who appear to be more likely than traditional laboratory participants to attempt to guess the desired interpretation behind experiments and to alter their behavior correspondingly (Berinsky et al. 2012). Concealing the true intent of a research study is a common method for experimental research in labor economics. For example, résumé audit studies routinely use deception by applying to job listings with fictional résumés. My experimental protocol was approved by the Duke Institutional Review Board. All participants were debriefed after completion of the study.

Each participant assessed ten résumés made up of two filler résumés and eight manipulated résumés. The eight manipulated résumés varied on sex, membership in an LGBT student group, and type of language used. The two filler résumés helped disguise the manipulation by using neutral adjectives (flexible, adaptable, talented, and reliable) and by reducing the proportion of résumés that are identifiable as an LGBT applicant. The two filler résumés were always presented first to the participant. The other eight résumés were presented in a random order.

The experiment over-represents LGBT résumés; however, among young people identification as LGBT or queer has dramatically increased. Approximately one-third of Generation Z (born between 1996 and 2003) self-identifies as bisexual or homosexual (Laughlin 2016). An estimated 7.3% of Millennials (born between 1980 and 1998) self-identify as LGBT (Gates 2017).

The participants were asked to view each résumé and then evaluate the job candidate on a number of characteristics. The survey was designed so that participants had to stay on each résumé page for a minimum of one minute. After viewing the whole résumé for one minute, the participants then rated the usefulness of the applicant’s work and extracurricular activities on pages that showed only that section of the résumé. Online Appendix B presents what the participant would see when rating the extracurricular activity and the slider the participant would use to rate the usefulness of the extracurricular activity.

The participant then evaluated the applicant’s personality, how strongly they would recommend the applicant, how willing the participant would be to work with the applicant, the applicant’s commitment to the job, recommended salary, and likelihood of success (based on the outcome measures in Bowles et al. 2007 and Correll, Benard, and Paik 2007). After rating all ten résumés, participants were asked to identify their own demographic information and answer questions about political ideology, including a question on the participant’s views toward LGBT people and on gender roles based on questions from the General Social Survey conducted by the National Opinion Research Center (NORC) at the University of Chicago.

There were 10 versions of the questionnaire, so that each résumé (made up of the work experience, education, and overall formatting) was paired once with each identity (the combination of sex, LGBT student group participation, and type of adjectives). For example, in one version of the questionnaire the first résumé was a man with masculine adjectives who was in an LGBT student group. In another version, the first résumé could be a woman with feminine adjectives who was not in an LGBT student group. Although each participant saw each résumé only once, each résumé was used with all of the manipulations over the 10 versions of the questionnaire. This experimental design allows for the inclusion of résumé fixed effects and participant fixed effects.

To increase the quality of the data analyzed, I used numerous methods to exclude participants who could be a computer program answering questions randomly or a person who was not paying attention to the survey. First, I set up the task on MTurk to allow responses only from those with high accuracy on previously submitted tasks on MTurk. Second, I incorporated an “attention check” question in the survey. The directions above the question instructed the participant to ignore the text of the question and to instead type a specific word in the text box. If a participant was clicking randomly or not reading the directions, they would not type the word into the text box. Seventy-nine participants failed to type the correct word in the text box and were excluded from the analyses. Third, I asked participants to indicate their sex in a text box; eight participants put their age in the text box instead of their sex and one put a series of nonsensical letters—these participants were excluded from the analyses. Fourth, I asked participants to indicate if the applicant was male, female, or indeterminate; 24 participants said the résumé was of indeterminate sex or incorrectly identified the applicant’s sex more than one time, so those participants were excluded. Finally, if the participant spent less than 26.2 minutes (the 5th percentile) on the survey, they were excluded; this affected 30 participants. Of the excluded participants, many failed more than one of the quality checks. In total, 878 participants passed all the quality checks. But note that the results shown in this article remain similar if all observations are included.

It is possible that participants became aware of the purpose of the experiment and altered their behavior (Berinsky et al. 2012). To check for this, I examined if participants rated the résumés they saw early in the group differently from how they rated the résumés they saw later. I detected no difference between how participants rated résumés viewed early in the group compared to those they viewed later. All patterns described in the results section are also found if analysis is restricted to the first four manipulated résumés viewed. This outcome suggests that even if participants became aware of the purpose of the experiment, they did not alter their behavior significantly.

Regression Framework

To test if participants show bias toward perceived LGBT résumés and if there is more bias toward male LGBT résumés than to female LGBT résumés, I first examine if participants rate résumés from LGBT men and women differently from non-LGBT applicants. I stratify by male and female participants. The outcome variable ( $y_{i, r}$ ) is the participant’s (r) assessment of each applicant’s (i) résumé.

\begin{matrix} y_{i, r} = α + β * I (LGBT activit y_{i}) + θ * I (Female resum e_{i}) \\ + \sum_{r = 1}^{n - 1} τ_{r} * I (participant = r) + \sum_{k = 1}^{7} ω_{k} * I (resum e_{i} = k) + η_{i, r} \end{matrix}

(1)

I then build on this base regression to examine if the LGBT effect is different for male and female résumés, or equivalently, if the female effect if different for LGBT and non-LGBT résumés.

\begin{matrix} y_{i, r} = α + β * I (LGBT activit y_{i}) + θ * I (Female resum e_{i}) \\ + δ * I (Female and LGBT activit y_{i}) + \\ + \sum_{r = 1}^{n - 1} τ_{r} * I (participant = r) + \sum_{k = 1}^{7} ω_{k} * I (resum e_{i} = k) + η_{i, r} \end{matrix}

(2)

Finally, I examine if the use of a masculine adjective affects the evaluation of the résumés because of the perceived sex and sexual orientation of the applicant. To do this, I augment Equation (1) to examine the difference within the four groups of how résumés with masculine adjectives are viewed compared to those with feminine adjectives.

\begin{matrix} y_{i, r} = α + β * I (LGBT activit y_{i}) + θ * I (Female resum e_{i}) \\ + φ I (Masculine adjective) + δ_{1} * I (Female and LGBT activit y_{i}) \\ + δ_{2} * I (Female and Masculine adjectiv e_{i}) \\ + δ_{3} * I (Masculine adjective and LGBT activit y_{i}) \\ + γ * I (Female and LGBT activity and Masculine adjectiv e_{i}) \\ + \sum_{r = 1}^{n - 1} τ_{r} * I (participant = r) + \sum_{k = 1}^{7} ω_{k} * I (resum e_{i} = k) + η_{i, r} \end{matrix}

(3)

Finite Mixture Model

To determine if the effects found in Equations (1), (2), and (3) are actually the average between two latent classes, I utilize finite mixture model (FMM) analysis. That is, perhaps one group of people do not have a reaction when women use masculine adjectives and another group has a strong reaction. The analysis in Equation (3) would identify only the weighted average between the two groups. FMM analysis examines if heterogeneity is in the sample based on unobserved characteristics.

Suppose there are two classes of participants, one with the relationship $y_{r, i} = x_{r, i} β_{1} + η_{r, i}$ and the other $y_{r, i} = x_{r, i} β_{2} + ω_{r, i}$ . Equation (4) shows the likelihood for participant r, who could be in the first latent class with probability $π_{1}$ or in the second class with probability $π_{2} = 1 - π_{1}$ ( $where π_{1} \in [0, 1]) .$ Participant r has I observations (I résumés reviewed by participant r) distributed with probability density function (pdf) f₁ if participant r is in group 1 and f₂ if they are in group 2. Assuming these I observations are independent, the product of their pdf represents the joint probability. All data are demeaned by participant to account for dependence between observations within the same participant. The sum of the two products, weighted by the probabilities of being in group 1 and group 2 is the likelihood function for person r.

\begin{matrix} L_{r} (π_{1}, π_{2}, β_{1}, σ_{1}^{2}, β_{2}, σ_{2}^{2} | y_{r}) = & π_{1} * Π_{i = 1}^{I} f_{1} (y_{r, i} | x_{r, i}, β_{1}, σ_{1}^{2}) \\ + π_{2} * Π_{i = 1}^{I} f_{2} (y_{r, i} | x_{r, i,} β_{2}, σ_{2}^{2}) \end{matrix}

(4)

If all N participants are independent and identically distributed, the likelihood function is

\begin{matrix} L (π_{1}, π_{2}, β_{1}, σ_{1}^{2}, β_{2}, σ_{2}^{2} | y) = \\ Π_{r = 1}^{N} [π_{1} * Π_{i = 1}^{I} f_{1} (y_{r, i} | x_{r, i}, β_{1}, σ_{1}^{2}) + π_{2} * Π_{i = 1}^{I} f_{2} (y_{r, i} | x_{r, i,} β_{2}, σ_{2}^{2})] \end{matrix}

(5)

Maximizing ln(L) by choosing $π_{1}, π_{2}, β_{1}, σ_{1}^{2}, β_{2}, σ_{2}^{2}$ will estimate a set of coefficients for each latent class and also the proportion of the sample that falls in each latent class. From this, I calculate the posterior probability that a participant is in a specific latent class, shown here for being in group 1:

\begin{matrix} \Pr (Respondent r \in Group 1) = \\ \frac{π_{1} * Π_{i = 1}^{I} f_{1} (y_{r, i} | x_{r, i}, β_{1}, σ_{1}^{2})}{π_{1} * Π_{i = 1}^{I} f_{1} (y_{r, i} | x_{r, i}, β_{1}, σ_{1}^{2}) + π_{2} * Π_{i = 1}^{I} f_{2} (y_{r, i} | x_{r, i,} β_{2}, σ_{2}^{2})} \end{matrix}

(6)

Results

Characteristics of Participants

The participants tend to be young and well-educated: More than 60% of the sample are less than age 35, and 50% have a bachelor’s degree or higher. The sample represents both men and women well, with 52.5% of the sample being female. The majority of participants (77%) are non-Hispanic white, but with sizable portions that are non-Hispanic African American (7%), Asian (5%), and multi-racial (4%). The vast majority of participants live in households with an income of $60,000 or less. The participants hold predominantly liberal views; 74% of participants agree or strongly agree that same-sex marriage should be legal.

Evaluating If the Treatment Was Salient to Participants

I first test if the use of adjectives in the objective statement was salient to the participant, that is, if the participant noticed the manipulation. I examine if the participants’ evaluation of the applicant’s personality is affected by the use of masculine adjectives relative to feminine adjectives. Participants evaluated 11 separate personality characteristics on how well they described the applicant (from 0 to 100). Figure 3 shows that within each sex by sexual orientation subgroup, the use of the masculine adjectives makes an applicant appear more confident and less kind. Although not shown, the same pattern holds for negative attributes; the use of the masculine adjectives makes an applicant appear less passive and more pushy. The patterns are the same for male and female participants.

Figure 3.

Participants’ Assessment of How “Confident” and “Kind” an Applicant Is

The results shown in Figure 3 suggest that the use of adjectives in the objective statement is effective—participants noticed and responded to the treatment. In each subgroup, the difference in the perceived personality characteristics (confident, kind, passive, and pushy) between the masculine and the feminine adjectives is statistically significant at the .001 level (robust standard errors, clustered at the participant level, with participant fixed effects). The use of masculine adjectives strongly and consistently affects how the participant views an applicant’s personality.

Participants frequently assigned applicants a score of zero for “pushy” and “passive”; to address this, I also evaluate dichotomous (positive or zero) versions of the “pushy” and “passive” measures in a logit model. Participants also frequently selected a multiple of 10 for their evaluations of “kind” and “confident”; I create 10 bins (from 0 to 9, 10 to 19, and so on) for the “kind” and “confident” measures that I evaluate in an ordered logit model. Results of the logit and ordered logit models mirror those described above and are all significant at the .001 level (robust standard errors, clustered at the participant level). The results for all 11 personality characteristics are statistically significant in each of the four subgroups and follow the same pattern; these results are available upon request.

Regression Results

Table 1 shows the results of Equations (1), (2), and (3), in which the recommended salary is regressed on indicator variables for LGBT activity and female résumé (columns (1) and (4)), then their interaction (columns (2) and (5)), and then adding interaction terms with the masculine adjectives (columns (3) and (6)). Table 1 shows the results for the “Salary” outcome. Online Appendix E contains the same regressions for the outcomes “Successful,”“Recommend,” and “Willing to work with.” All three outcomes have similar patterns to the “Salary” outcome.

Table 1.

Results of an OLS Regression of “Salary”

	Salary			Salary
	Male participants			Female participants
	(1)	(2)	(3)	(4)	(5)	(6)
LGBT activity on résumé	−2.139***	−2.508***	−3.125***	−0.729*	−0.406	−0.600
	(0.529)	(0.681)	(0.975)	(0.433)	(0.581)	(0.822)
Female résumé	1.210**	0.841	2.164**	0.412	0.736	1.019
	(0.498)	(0.632)	(0.948)	(0.457)	(0.624)	(0.842)
LGBT and Female		0.738	0.0491		−0.646	−0.743
		(0.895)	(1.309)		(0.768)	(1.132)
Masculine adjective on résumé			1.211			0.475
			(0.980)			(0.879)
LGBT and Masculine adjective			1.230			0.388
			(1.292)			(1.199)
Masculine adjective and Female résumé			−2.647**			−0.568
Masculine adjective and Female résumé			(1.330)			(1.186)
LGBT and Female résumé and Masculine adjective			1.383			0.194
LGBT and Female résumé and Masculine adjective			(1.903)			(1.759)
Constant	46.65***	46.83***	46.19***	45.82***	45.66***	45.42***
	(0.707)	(0.710)	(0.864)	(0.766)	(0.796)	(0.913)
Observations	3,328	3,328	3,328	3,688	3,688	3,688
R-squared	0.476	0.476	0.478	0.481	0.481	0.481

Notes: Robust standard errors in parentheses. Controls include participant and résumé fixed effects. Errors are robust and clustered at the participant level. Outcome variable could take on values from 0 to 100. OLS, ordinary least squares.

***

p < 0.01; **p < 0.05; *p < 0.1.

As shown in Table 1, column (1), male participants rate résumés with an LGBT activity more negatively than they rate identical résumés with a non-LGBT activity, and they rate female résumés slightly more positively than male résumés. Column (2) shows the negative effect of an LGBT activity is slightly smaller for female LGBT résumés, but still significant (−2.508+.738 = −1.77, p value of F test is .0125). Note that LGBT female résumés are not statistically different from non-LGBT male résumés (−2.508+.738+.841 = −.929, p value of F test is .196).

Column (3) tests the impact of using a masculine adjective. Among the omitted category (non-LGBT and male), using a masculine adjective causes participants to suggest a slightly higher salary, although this is not statistically significant. Female résumés, however, experience a more negative effect from using a masculine adjective than male résumés experience. For female résumés with an LGBT activity and using a masculine adjective, this negative effect does not occur (−2.647+1.383 = −1.264, p value of F test is .3428).

An important note is that women who use feminine language are rated better on these measures than any other group, including men of both adjective types. This result is consistent with Bowles et al. (2007), who found that women who did not negotiate were rated more highly than men who negotiated and men who did not negotiate. What is considered a negative reaction to women using masculine adjectives could also be viewed as a premium for women using feminine adjectives. To examine this more closely, I compare male and female applicants among the filler résumés (the first two résumés the participants viewed). These filler résumés used gender-neutral adjectives and the non-LGBT activity. Female filler résumés were also rated more highly than male filler résumés on salary and how successful the applicant would be (difference in means are significant at the .05 level, with participant fixed effects and robust clustered errors at the participant level). This finding suggests a preference for female applicants except when they use masculine adjectives, rather than a preference for women who use feminine adjectives over all other groups. Female applicant preference is consistent with literature on prosocial behavior, in which male participants help women more than men (Eagly and Crowley 1986; Eagly 2009). Such analysis must be interpreted cautiously, however, because the filler résumés were intended to help the participant adjust to the experiment rather than be used in the analysis.

For the most part, female participants did not have a statistically significant difference in their evaluations. They rate LGBT résumés slightly worse than the non-LGBT résumés in the base regression (column (4)) but not in the interacted regressions. Female participants show no difference in how they rate résumés that use masculine or feminine adjectives for any sex by sexual orientation combination. Female participants largely do not rate résumés dissimilarly based on adjective use, except for showing less willingness to work with people who use masculine adjectives. This is not different based on sex or perceived sexual orientation. Compared to male participants, female participants are more willing to work with non-LGBT women applicants, but this does not diminish when they use masculine adjectives.

LGBT Activities on Résumés: A Deeper Look

To examine if only the final outcomes are affected by the inclusion of an LGBT activity, I also examine if the LGBT extracurricular activities are perceived as less useful than equivalent activities. To control for the value of extracurricular activities, I analyze a particular résumé on which the extracurricular résumé entry for the non-LGBT group is identical to the LGBT group, except for the name. The first bullet point in the job description states “Planned and organized events that promoted diversity and raised awareness on various topics.” The entry includes other details to demonstrate the magnitude of the role, such as “managed a committee of 10 to 12 members” (see Online Appendix B for full text). The LGBT club was named the “LGBT Alliance,” and the non-LGBT club was named “Student Activities Board.”

After viewing the whole résumé for one minute, participants were asked to evaluate the usefulness of the applicant’s extracurricular activities on a scale of 1 to 10 on a page that displayed only the participant’s name, the objective statement, and the extracurricular activity. Despite having the exact detailed description of the role, the résumés with the non-LGBT club were rated as 2.84 on the useful scale, and the LGBT version was rated 2.51 (p = .06, robust standard errors). Because the usefulness ratings took on only whole numbers, I also use an ordered logit model to examine this question. The odds ratio on the LGBT activity indicator is 0.75 (p = .03, robust standard errors; N = 702), indicating again that the LGBT extracurricular activity is viewed as less useful than the identical entry for a non-LGBT group. Restricting to only those participants who agree or strongly agree that same-sex marriage should be legal also results in a statistically significant difference: 2.96 for the non-LGBT group and 2.46 for the LGBT group (p = .02) and an odds ratio of 0.68 (p = .01).

This negative association leaks into the assessment of the applicant’s work history for male applicants being evaluated by male participants. After viewing the whole résumé for one minute, the participant is shown a page that lists only the applicant’s name and work history (no extracurricular activity) and asked to rate the usefulness of the applicant’s work history on a scale of 1 to 10. Recall that the following regression includes résumé fixed effects—the résumés have identical work experiences. Table 2, column (2) shows that for male applicants evaluated by male participants, having an LGBT activity on the résumé results in a lower assessment of the usefulness of the applicant’s work history (−0.203). Note that female résumés with an LGBT activity do not have this effect (−.203+.243 = .04, p value = .6723). Column (3) shows that LGBT résumés that use masculine adjectives partially ameliorate this effect (−.308+.211 = .097, p value of F test .4263). Columns (4), (5), and (6) show that female participants did not have this negative reaction to the work history on male résumés with an LGBT activity.

Table 2.

Results of an OLS Regression of “Usefulness of work history” in Equations (1), (2), and (3)

	Usefulness of work history			Usefulness of work history
	Male participants			Female participants
	(1)	(2)	(3)	(4)	(5)	(6)
LGBT activity on résumé	−0.0816	−0.203**	−0.308**	−0.0345	−0.0795	−0.0601
	(0.0667)	(0.0907)	(0.132)	(0.0658)	(0.0945)	(0.129)
Female résumé	0.140*	0.0192	0.151	0.00458	−0.0404	0.117
	(0.0744)	(0.0996)	(0.133)	(0.0667)	(0.0908)	(0.125)
LGBT and Female		0.243*	0.219		0.0899	0.0424
		(0.127)	(0.177)		(0.126)	(0.171)
Masculine adjective on résumé			0.0134			0.204
Masculine adjective on résumé			(0.127)			(0.127)
LGBT and Masculine adjective			0.211			−0.0389
LGBT and Masculine adjective			(0.179)			(0.187)
Masculine adjective and Female résumé			−0.265			−0.314*
Masculine adjective and Female résumé			(0.181)			(0.177)
LGBT and Female résumé and Masculine adjective			0.0483			0.0952
LGBT and Female résumé and Masculine adjective			(0.252)			(0.255)
Constant	5.211***	5.271***	5.263***	5.546***	5.569***	5.467***
	(0.112)	(0.117)	(0.127)	(0.116)	(0.122)	(0.136)
Observations	3,328	3,328	3,328	3,688	3,688	3,688
R-squared	0.591	0.592	0.593	0.584	0.584	0.585

Notes: Robust standard errors in parentheses, clustered by participant. Controls include participant and résumé fixed effects. Errors are robust and clustered at the participant level. Usefulness variable could take on values from 0 to 10. OLS, ordinary least squares.

***

p < 0.01; **p < 0.05; *p < 0.1.

Table 3 shows that both male and female participants view LGBT résumés, both male and female résumés, as more “Pushy” than résumés with the non-LGBT activity.

Table 3.

Results of an OLS Regression of “Pushy” in Equations (1), (2), and (3)

	Pushy			Pushy
	Male participants			Female participants
	(1)	(2)	(3)	(4)	(5)	(6)
LGBT activity on résumé	2.212***	1.820**	1.978**	1.945***	1.811**	0.927
	(0.519)	(0.726)	(0.840)	(0.543)	(0.716)	(0.605)
Female résumé	−0.998**	−1.389**	−0.561	−0.925*	−1.059	0.346
	(0.491)	(0.634)	(0.714)	(0.483)	(0.649)	(0.579)
LGBT and Female		0.783	−0.577		0.267	0.285
		(0.937)	(1.063)		(0.893)	(0.907)
Masculine adjective on résumé			7.816***			10.34***
			(1.010)			(1.084)
LGBT and Masculine adjective			−0.331			1.758
			(1.417)			(1.452)
Masculine adjective and Female résumé			−1.652			−2.808**
Masculine adjective and Female résumé			(1.285)			(1.285)
LGBT and Female résumé and Masculine adjective			2.720			−0.0336
LGBT and Female résumé and Masculine adjective			(1.952)			(1.908)
Constant	16.02***	16.21***	12.20***	11.80***	11.86***	6.664***
	(0.764)	(0.780)	(0.846)	(0.821)	(0.847)	(0.854)
Observations	3,328	3,328	3,328	3,688	3,688	3,688
R-squared	0.553	0.553	0.589	0.420	0.420	0.489

Notes: Robust standard errors in parentheses, clustered by participant. Controls include participant and résumé fixed effects. Errors are robust and clustered at the participant level. Pushy variable could take on values from 0 to 100. OLS, ordinary least squares.

***

p < 0.01; **p < 0.05; *p < 0.1.

Taken together, the above results suggest that having an LGBT-related extracurricular activity is viewed negatively, with the negative effect being most consistent for male LGBT résumés being evaluated by male participants. LGBT résumés, both male and female, were seen as more pushy than non-LGBT résumés. Male LGBT résumés were rated more negatively on the multiple hireability questions, particularly by male participants. Their extracurricular activity was viewed as less useful than an identical activity that is not LGBT related; this remained true even among those who support same-sex marriage. Moreover, when men evaluated the usefulness of a male applicant’s work history on a separate page from the extracurricular activity, they rated the work experience of a male applicant who has the LGBT activity as less useful relative to the identical work history of another applicant without the LGBT activity. The spillover did not occur for female applicants or for female participants.

Direct Measure of LGBT Discrimination

An alternative method to examine LGBT discrimination is to look directly at how each participant rates résumés with the LGBT activity compared to résumés with the non-LGBT activity. I first regress the outcome variable y on participant and résumé fixed effects.

y_{i, r} = β + \sum_{r = 1}^{n - 1} τ_{r} * I (participant = r) + \sum_{k = 1}^{7} ω_{k} * I (resum e_{i} = k) + η_{i, r}

I then calculate the residual for each observation:

e_{i, r} = y_{i, r} - {\hat{y}}_{i, r}

I calculate ${\hat{δ}}_{r}$ , the difference in the average residual between the résumés with the non-LGBT activity and the LGBT activity:

{\hat{δ}}_{r} = \frac{\sum_{i = 1}^{8} e_{i, r} I (LGBT = 0) I (participant = r)}{4} - \frac{\sum_{i = 1}^{8} e_{i, r} I (LGBT = 1) I (participant = r)}{4}

I then average ${\hat{δ}}_{r}$ across the four hireability measures.

A ${\hat{δ}}_{r}$ of 0 reflects no discrimination, a positive value shows that non-LGBT résumés were rated better than were résumés with the LGBT activity, after accounting for résumé and participant fixed effects. On average, male participants had a ${\hat{δ}}_{r}$ of 3.06 and female participants had a ${\hat{δ}}_{r}$ of 1.35, indicating that on average participants rated résumés with the non-LGBT activity higher than the LGBT résumés. Of all male participants, 60.3% had a positive ${\hat{δ}}_{r},$ and 53.4% of female participants had a positive ${\hat{δ}}_{r}$ .

As shown in Figure 4, participants who hold more conservative views on gender, sexual orientation, and gun laws have larger differences between résumés regarding the non-LGBT activity and the LGBT activity.

Figure 4.

Measure of LGBT Discrimination Based on Participant Political Beliefs

As shown in Figure 5, older participants had larger measures of LGBT discrimination. This finding is important because the participants in this sample are disproportionately young and liberal; the results shown in Figures 4 and 5 suggest that the overall discrimination against résumés with an LGBT activity are muted.

The participant’s education was not associated with differences in LGBT discrimination; most education groups had moderate levels of discrimination. Participants with less than high school showed less discrimination, but this group comprised only six participants. Similarly, there were small differences by race, with black participants demonstrating slightly more discrimination. Participants who earn less than $30,000 per year showed less discrimination than did wealthier participants.

Figure 5.

Measure of LGBT Discrimination Based on Self-Reported Participant Demographics

Magnitude of Effect

To assess the importance of the magnitude of the negative reaction to an LGBT activity, Table 4 mimics a theoretical selection of résumés for which an employer chooses to interview the applicants in the top 25% of applicants. I first calculate the 75th percentile of the four outcome variables for non-LGBT résumés (demeaned by participant) for male participants. I then calculate the proportion of résumés that fall above the 75th percentile. As shown in the table, substantially fewer LGBT résumés fall into the top 25% for all four outcome variables.

Table 4.

Proportion of Résumés That Fall in Top 25% for Male Participants

		Non-LGBT activity (%)	LGBT activity (%)
Male résumés	Recommend	22.1	17.9
	Salary	23.9	18.6
	Willing to work with	22.2	13.7
	Successful	22.1	17.8
Female résumés	Recommend	27.5	21.5
	Salary	25.7	24.4
	Willing to work with	27.5	19.0
	Successful	27.2	22.4

Finite Mixture Model

To investigate if the penalty for including an LGBT activity on a résumé is the average between two latent classes, I apply the FMM from the likelihood function in Equation (5). This analysis is shown in Table 5 and reveals two distinct classes among male participants, but not among female participants. In the following regressions, all the data are demeaned by participant to account for the dependence between observations within the same participant.

Table 5.

Results of FMM Model Regression Examining Effect of Including an LGBT Activity on Suggested Salary

	Salary		Salary
	Male participants		Female participants
	Class 1	Class 2	Class 1	Class 2
LGBT résumés	−0.241	−2.967***	−0.99**	−0.455
	(0.384)	(0.999)	(0.430)	(0.977)
Female résumé	−0.130	1.345	−0.491	1.478
	(0.397)	(0.996)	(0.431)	(0.978)
LGBT female résumé	0.231	1.091	0.457	−1.670
	(0.552)	(1.410)	(0.585)	(1.380)
Observations	3,328	3,328	3,688	3,688
Estimated proportion	.36	.64	.39	.61

Notes: Standard errors in parentheses. All variables are demeaned by participant. Outcome variables can take on values from 0 to 100. FMM, finite mixture model.

***

p < 0.01; **p < 0.05; *p < 0.1.

Online Appendix F shows the pattern is similar for “Successful” and “Willing to work with”—for both outcome variables a large group of male participants have a strong negative LGBT effect and a strong preference to work with women. Both classes among “Recommend” have a similar negative reaction. Among female participants, neither class has a statistically significant negative LGBT effect, although a strong preference was shown to work with women without the LGBT activity, which is eliminated for women with an LGBT activity.

Using the FMM model to examine the negative reaction to using masculine adjectives also reveals two latent classes. As shown in Table 6, among male participants, a large latent class had a strong negative response when non-LGBT women use masculine adjectives and a strong positive response when they use feminine adjectives. This latent class also tends to have stronger negative reactions to LGBT male résumés. Results for “Successful,”“Willing to work with,” and “Recommend” are again in Online Appendix F and all have similar patterns to “Salary.”

Table 6.

Results of FMM Model Examining Effect of Including a Masculine Adjective on Suggested Salary (All Resumes)

	Salary		Salary
	Male participants		Female participants
	Class 1	Class 2	Class 1	Class 2
LGBT activity on résumé	0.338	−3.870***	−0.934	−1.073
	(0.569)	(1.434)	(0.602)	(1.335)
Female résumé	0.824	3.023**	−0.676	1.888
	(0.596)	(1.424)	(0.616)	(1.339)
LGBT and Female	−1.267	0.421	−0.211	−1.258
	(0.822)	(2.020)	(0.842)	(1.890)
Masculine adjective on résumé	0.695	1.623	−0.121	0.762
	(0.592)	(1.426)	(0.611)	(1.338)
LGBT and Masculine adjective	−1.119	1.712	−0.211	1.252
	(0.825)	(2.022)	(0.854)	(1.886)
Female and masculine adjective	−1.788**	−3.386*	0.190	−0.860
	(0.813)	(2.013)	(0.876)	(1.892)
LGBT résumé & female & masculine adjectives	2.836**	1.453	1.360	−0.681
	(1.140)	(2.851)	(1.236)	(2.672)
Observations	3,328	3,328	3,688	3,688
Estimated proportion	.36	.64	.37	.63

Notes: Standard errors in parentheses. All variables are demeaned by participant. Outcome variables can take on values from 0 to 100. FMM, finite mixture model.

***

p < 0.01; **p < 0.05; *p < 0.1.

Among female participants, a negative reaction to using masculine adjectives for “Willing to work” is observed, but this does not vary by the sex or sexual orientation of the applicant. One group among female participants shows a strong preference to work with non-LGBT women who do not use the masculine adjective. But beyond that, female participants show no evidence of having a differential reaction to applicant’s use of masculine language even when split into latent classes.

Posterior Probabilities

Using Equation (6) to calculate posterior probabilities shows that the male participants are distinctly split between the two latent classes. That is, almost all participants have posterior probability of being in the high discrimination groups of either more than .9 or less than .1; there are few ambiguous participants. More information on the posterior probabilities for each outcome variable for all FMM models is available upon request.

The two types of discrimination—against men with an LGBT organization and non-LGBT women who use masculine adjectives—are strongly correlated. Table 6 showed that discrimination against LGBT male résumés was higher in Class 2, which had stronger discrimination against non-LGBT women with masculine adjectives. To look at this more directly, I run a simplified version of the FMM model examining the impact of masculine adjectives that compares only men and women without the LGBT activity. The results for “Salary” are shown in Table 7, and the results for “Willing to work with,”“Recommend,” and “Successful” are shown in Online Appendix F.

Table 7.

Results of FMM Model Examining Effect of Including a Masculine Adjective on Suggested Salary (Non-LGBT Resumes)

	Salary
	Male participants
	Class 1	Class 2
Female résumé	0.324	3.154**
	(0.499)	(1.380)
Masculine adjective on résumé	0.0536	1.892
	(0.495)	(1.379)
Female résumé and a masculine adjective	−1.206*	−3.590*
	(0.712)	(1.951)
Observations	1,664	1,664

Notes: Standard errors in parentheses. FMM, finite mixture model.

***

p < 0.01; **p < 0.05; *p < 0.1.

Table 8 shows the correlation between the posterior probability of being in the higher discrimination group for the LGBT model and for the masculine adjective model when estimated on non-LGBT résumés. For the “Salary” outcome, Table 8 shows the correlation between the posterior probability of being in Class 2 on Table 5 and Class 2 in Table 7.

Table 8.

Correlation between Posterior Probability of Being in Higher Discrimination Group for the LGBT Model (Table 5) and the Masculine Adjective Model (Table 7)

Recommend	Successful	Salary	Willing to work with
.9235	.9977	.9377	.9846

All four variables have a very high correlation between the two high discrimination groups. This outcome suggests that men who hold strong beliefs that women should act in a feminine manner also have the stronger negative reaction to men with an LGBT activity. These are two different threats to social norms that both affect how these men evaluate a job applicant’s résumé.

Conclusions

Gay men and lesbian women have labor market outcomes that differ from those of similarly educated heterosexual people: Gay men earn less than heterosexual men, and lesbians earn more than heterosexual women. In this article, I examine if social norms and cognitive biases affect how LGBT men and women are evaluated by potential employers. I test interrelated questions that may explain why labor market outcomes differ between gay men and lesbians. First, I examine if participants show bias toward perceived-LGBT résumés and if the bias is greater toward male résumés with an LGBT activity than toward female résumés with an LGBT activity. Second, I assess if the same behavioral prescriptions that face heterosexual women apply equally to LGBT women. I then test if the first two questions are driven by male or female participants.

I find that résumés with an LGBT activity are penalized. These résumés are viewed more negatively on numerous personality characteristics and have lower ratings on the hireability measures. These effects are strongest for perceived-LGBT men being evaluated by male participants. A perceived-LGBT man’s work history is viewed as less useful when compared to a résumé with an identical work history, suggesting that the negative view about an LGBT activity on a man’s résumé spills over to the evaluation of other unrelated aspects of the résumé. This pattern held even among laboratory participants who reported progressive views on LGBT rights. The negative evaluation of résumés with an LGBT activity are strongest among older participants and those with more conservative political beliefs. These findings suggest there is a substantial and persistent penalty for including an LGBT activity on a résumé.

But this study disclosed more than a penalty for including an LGBT activity—I also show that women with an LGBT activity on their résumé are not held to the same behavioral norms that perceived-heterosexual women face. Numerous laboratory studies have shown that men prefer when women act in traditionally feminine ways rather than in traditionally masculine ways (Rudman 1998; Rudman and Glick 1999, 2001; Gill 2004; Heilman et al. 2004; Heilman and Chen 2005; Bowles et al. 2007). I find that using feminine language is viewed positively when used by non-LGBT women, but not when used by women with an LGBT activity on their résumé.

Male participants rated non-LGBT female applicants who use masculine adjectives unfavorably relative to when they use feminine adjectives. The difference between male and female applicants who use masculine or feminine language is significant for three important hireability measures: how much the participant would recommend the company hire the applicant, the recommended starting salary, and their likelihood of success. By contrast, perceived-LGBT women are exempt from this effect. Perceived-LGBT female applicants are not rated differently when they use masculine adjectives relative to when they use feminine adjectives. This finding is striking—it is not just an anti-LGBT effect, but rather that perceived-LGBT women are held to different behavioral norms than are perceived-heterosexual women.

Latent class analysis suggests that this average effect is driven by a larger effect in a majority of male participants tempered by a weaker reaction in a minority of male participants. Discrimination is common and widespread, not the extreme reaction of a few bad actors. The same men who reacted negatively to LGBT activities on men’s résumés also reacted the most strongly to the difference between feminine and masculine adjectives on non-LGBT women’s résumés.

Female participants did not rate any group differently from the others based on the applicant’s choice of adjective. Even when split into latent classes, both classes of female participants rated non-LGBT female applicants who used masculine language equally to those applicants who used feminine adjectives. For the most part, female participants also did not rate a group differently from the others if they included an LGBT activity, although there was a preference to work with non-LGBT women.

Taken together, these results suggest social norms and cognitive biases that affect how gender and sexual orientation are perceived by male decision makers in employment settings. Literature from psychology has found persistent patterns of cognitive biases that affect how women and sexual minorities are perceived—and indeed these affected how laboratory participants rated résumés. Two distinct violations of norms, heterosexual women acting in traditionally masculine ways and an LGBT identity, both negatively affect how the majority of men evaluate a job applicant’s résumé.

These results are consistent with patterns of occupational segregation and earnings differences. Gay men are over-represented in female-dominated fields, which could be driven by women engaging in less discrimination against LGBT men. Perceived-LGBT women benefit in some ways by being exempt from norms facing heterosexual women, and this may allow gay women to thrive in male-dominated fields.

These results are inconsistent with the theory that a driving factor is information about productivity. If the impetus was a concern for productivity, male and female participants should have similar reactions to the adjectives. For productivity to result in only male participants evaluating résumés differently based on the adjectives used and only for résumés from heterosexual women, male and female participants would have to perceive dramatically different informational content in adjectives used specifically by heterosexual women.

Supplemental Material

DS_10.1177_0019793919832273 – Supplemental material for Gender, Sexual Orientation, and Behavioral Norms in the Labor Market

Supplemental material, DS_10.1177_0019793919832273 for Gender, Sexual Orientation, and Behavioral Norms in the Labor Market by Marina Mileo Gorsuch in ILR Review

Footnotes

Acknowledgements

I thank Duke University for generously providing funding for this project. This project has also benefited from insightful feedback from many people, including participants at the Pink Papers: LGBT Research session at the American Economic Association (AEA) organized by Kitt Carpenter. I also received invaluable feedback from David Eisnitz, Seth Sanders, Amar Hamoudi, Liz Ananat, and Ken Dodge.

Supplemental Online Appendices are available at .

For information regarding the data and/or computer programs used for this study, please address correspondence to mmgorsuch952@stkate.edu.

References

Ahmed

Ali M.

Andersson

Lina

Hammarstedt

Mats

. 2013. Are gay men and lesbians discriminated against in the hiring process? Southern Economic Journal 79(3): 565–85.

Amanatullah

Emily T.

Morris

Michael W.

2010. Negotiating gender roles: Gender differences in assertive negotiating are mediated by women’s fear of backlash and attenuated when negotiating on behalf of others. Journal of Personality and Social Psychology 98(2): 256–67.

Antecol

Heather

Jong

Anneke

Steinberger

Michael D.

2008. Sexual orientation wage gap: The role of occupational sorting and human capital. Industrial and Labor Relations Review61(4): 518–43.

Badgett

Lee

. 1995. The wage effects of sexual orientation. Industrial and Labor Relations Review 48(4): 726–39.

Berg

Nathan

Lien

Daniel

. 2002. Measuring the effect of sexual orientation on income: Evidence of discrimination? Contemporary Economic Policy 20(4): 394–414.

Berinsky

Adam

Huber

Gregory

Lenz

Gabriel

. 2012. Evaluating online labor markets for experimental research: Amazon.com’s Mechanical Turk. Political Analysis 20(3): 351–68.

Bertrand

Marianna

Mullainathan

Sendhil

. 2004. Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review 94(4): 991–1013.

Black

Dan

Gates

Gary

Sanders

Seth

Taylor

Lowell J.

2000. Demographics of the gay and lesbian population in the United States: Evidence from available systematic data sources. Demography 37(2): 139–54.

Black

Dan

Makar

Hoda R.

Sanders

Seth G.

Taylor

Lowell J.

2003. The earnings effect of sexual orientation. Industrial and Labor Relations Review 56(3): 449–69.

10.

Blandford

John

. 2003. The nexus of sexual orientation and gender in the determination of earnings. Industrial and Labor Relations Review 56(4): 622–42.

11.

Blau

Francine D.

Kahn

Lawrence M.

2017. The gender wage gap: Extent, trends, and explanations. Journal of Economic Literature 55(3): 789–865.

12.

Bowles

Hannah

Babcock

Linda

Lai

Lei

. 2007. Social incentives for gender differences in the propensity to initiate negotiations: Sometimes it does hurt to ask. Organizational Behavior and Human Decision Processes 103(1): 84–103.

13.

Broverman

Inge K.

Vogel

Susan Raymond

Broverman

Donald M.

Clarkson

Frank E.

Rosenkrantz

Paul S.

1972. Sex-role stereotypes: A current appraisal. Journal of Social Issues 28(2): 59–78.

14.

Buhrmester

Michael

Kwang

Tracy

Gosling

Samuel D.

2011. Amazon’s Mechanical Turk: A new source of inexpensive, yet high-quality, data? Perspectives on Psychological Science 6(1): 3–5.

15.

Carpenter

Christopher S.

2005. Self-reported sexual orientation and earnings: Evidence from California. Industrial and Labor Relations Review 58(2): 258–73.

16.

Carpenter

Christopher S.

2007. Revisiting the income penalty for behaviorally gay men: Evidence from NHANES III. Labour Economics 14(1): 25–34.

17.

Carpenter

Christopher S.

2008. Sexual orientation, income, and non-pecuniary economic outcomes: New evidence from young lesbians in Australia. Review of Economics of the Household 6(4): 391–408.

18.

Carrington

Christopher

. 1999. No Place Like Home: Relationships and Family Life among Lesbians and Gay Men. Chicago: University of Chicago Press.

19.

Census Bureau. 2012. File A. Genealogy data: Frequently occurring surnames from Census 2000. Accessed at http://www.census.gov/genealogy/www/data/2000surnames/ (January 2014).

20.

Correll

Shelley J.

Benard

Stephen

Paik

. 2007. Getting a job? Is there a motherhood penalty? American Journal of Sociology 112(5): 1297–1339.

21.

Cushing-Daniels

Brendan

Yeung

Tsz-Ying

. 2009. Wage penalties and sexual orientation: An update using the General Social Survey. Contemporary Economic Policy 27(2): 164–75.

22.

Deaux

Kay

Lewis

Laurie L.

1984. The structure of gender stereotypes: Interrelationships among components and gender label. Journal of Personality and Social Psychology 46(5): 991–1004.

23.

Eagly

Alice

. 2009. The his and hers of prosocial behavior: An examination of the social psychology of gender. American Psychologist 64(8): 644–58.

24.

Eagly

Alice

Crowley

Maureen

. 1986. Gender and helping behavior: A meta-analytic review of the social psychology literature. Psychology Bulletin 100(3): 283–308.

25.

Elmslie

Bruce

Tebaldi

Edinaldo

. 2007. Sexual orientation and labor market discrimination. Journal of Labor Research 28(3): 436–53.

26.

Frank

Jeff

. 2006. Gay glass ceilings. Economica 73(291): 485–508.

27.

Gates

Gary J

. 2017. In U.S., more adults identifying as LGBT. Gallup Social and Policy Issues. Accessed at http://news.gallup.com/poll/201731/lgbt-identification-rises.aspx (January 2018).

28.

Gill

Michael J.

2004. When information does not deter stereotyping: Prescriptive stereotyping can foster bias under conditions that deter descriptive stereotyping. Journal of Experimental Social Psychology 40(5): 619–32.

29.

Gough

Brendan

. 2002. “I’ve always tolerated it but . . .”: Heterosexual masculinity and the discursive reproduction of homophobia. In Coyle

Adrian

Kitzinger

Celia

(Eds.), Lesbian and Gay Psychology: New Perspectives, pp. 219–238. Oxford, UK: BPS Blackwell.

30.

Halberstam

Jack

. 1998. Female masculinity. Durham, NC: Duke University Press.

31.

Heilman

Madeline E.

1995. Sex stereotypes and their effects in the workplace: What we know and what we don’t know. Journal of Social Behavior & Personality 10(6): 3–26.

32.

Heilman

Madeline E.

2001. Description and prescription: How gender stereotypes prevent women’s ascent up the organizational ladder. Journal of Social Issues 57(4): 657–74.

33.

Heilman

Madeline E.

Chen

Julie J.

2005. Same behavior, different consequences: Reactions to men’s and women’s altruistic citizenship behavior. Journal of Applied Psychology 90(3): 431–41.

34.

Heilman

Madeline E.

Parks-Stamm

Elizabeth J.

2007. Gender stereotypes in the workplace: Obstacles to women’s career progress. Advances in Group Processes 24: 47–77.

35.

Heilman

Madeline E.

Block

Caryn J.

Martell

Richard F.

1995. Sex stereotypes: Do they influence perceptions of managers? Journal of Social Behavior and Personality 10(6): 237–52.

36.

Heilman

Madeline E.

Wallen

Aaron S.

Fuchs

Daniella

Tamkins

Melinda M.

2004. Penalties for success: Reactions to women who succeed at male gender-typed tasks. Journal of Applied Psychology 89(3): 416–27.

37.

Herek

Gregory M.

2000. Sexual prejudice and gender: Do heterosexuals’ attitudes toward lesbians and gay men differ? Journal of Social Issues 56(2): 251–66.

38.

Herek

Gregory M.

2002. Gender gaps in public opinion about lesbians and gay men. Public Opinion Quarterly 66(1): 40–66.

39.

Horton

John J.

Rand

David G.

Zeckhauser

Richard J.

2011. The online laboratory: Conducting experiments in a real labor market. Experimental Economics 14(3): 399–425.

40.

Jepsen

Lisa K.

2007. Comparing the earnings of cohabiting lesbians, cohabiting heterosexual women, and married women: Evidence from the 2000 Census. Industrial Relations 46(4): 699–727.

41.

Kite

Mary E.

Deaux

Kay

. 1987. Gender belief systems: Homosexuality and the implicit inversion theory. Psychology of Women Quarterly 11(1): 83–96.

42.

Kite

Mary E.

Whitley

Bernard E.

Jr. , Jr. 1996. Sex differences in attitudes toward homosexual persons, behaviors, and civil rights: A meta-analysis. Personality and Social Psychology Bulletin 22(4): 336–53.

43.

Klawitter

Marieka

. 2011. Multilevel analysis of the effects of antidiscrimination policies on earnings by sexual orientation. Journal of Policy Analysis and Management 30(2): 334–58.

44.

Klawitter

Marieka

. 2015. Meta-analysis of the effects of sexual orientation on earnings. Industrial Relations 54(1): 4–32.

45.

Laughlin

Shepard.

2016. Gen Z goes beyond gender binaries in new Innovation Group data. J. Walter Thompson Intelligence. Accessed at https://www.jwtintelligence.com/2016/03/gen-z-goes-beyond-gender-binaries-in-new-innovation-group-data/ (January 2018).

46.

Levitt

Steven D.

Dubner

Stephen J.

2005. Freakonomics. New York: Harper Collins.

47.

Martell

Michael E.

2012. Differences do not matter: Exploring the wage gap for same-sex behaving men. Eastern Economic Journal 39(1): 45–71.

48.

Moskowitz

David A.

Rieger

Gerulf

Roloff

Michael E.

2010. Heterosexual attitudes toward same-sex marriage. Journal of Homosexuality 57(2): 325–36.

49.

National Center for Science and Engineering Statistics. 2013. Table 11. Women as a percentage of all bachelor’s recipients, by major field group: 1966–2010. Science and Engineering Degrees: 1966–2010. Alexandria, VA: NCSES, National Science Foundation.

50.

Pedulla

David S.

2014. The positive consequences of negative stereotypes: Race, sexual orientation, and the job application process. Social Psychology Quarterly 77(1): 75–94.

51.

Proschan

Frank

. 1997. Recognizing gay and lesbian speech. American Anthropologist 99(1): 164–66.

52.

Raja

Sheela

Stokes

Joseph P.

1998. Assessing attitudes toward lesbians and gay men: The modern homophobia scale. International Journal of Sexuality and Gender Studies 3(2): 113–34.

53.

Rudman

Laurie A.

1998. Self-promotion as a risk factor for women: The costs and benefits of counterstereotypical impression management. Journal of Personality and Social Psychology 74(3): 629–45.

54.

Rudman

Laurie

Glick

Peter

. 1999. Feminized management and backlash toward agentic women: The hidden costs to women of a kinder, gentler image of middle managers. Journal of Personality and Social Psychology 77(5): 1004–10.

55.

Rudman

Laurie

Glick

Peter

. 2001. Prescriptive gender stereotypes and backlash toward agentic women. Journal of Social Issues 57(4): 732–62.

56.

Rudman

Laurie

Phelan

Julie E.

2008. Backlash effects for disconfirming gender stereotypes in organizations. Research in Organizational Behavior 28: 61–79.

57.

Sabia

Joseph J.

2014. Sexual orientation and wages in young adulthood: New evidence from Add Health. Industrial and Labor Relations Review 67(1): 239–67.

58.

Tilcsik

András

. 2011. Pride and prejudice: Employment discrimination against openly gay men in the United States. American Journal of Sociology 117(2): 586–626.

59.

Weichselbaumer

Doris

. 2003. Sexual orientation discrimination in hiring. Labour Economics 10(6): 630–42

60.

Weichselbaumer

Doris

. 2004. Is it sex or personality? The impact of sex stereotypes on discrimination in applicant selection. Eastern Economic Journal 30(2): 159–86.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.52 MB