Heterosexism in the Workplace: Do Race and Gender Matter?

Abstract

We examined the influence of race and gender on reports of sexual orientation discrimination and the decision to disclose a gay identity at work.  Our national sample was comprised of 534 gay and lesbian employees, 162 of whom were gay people of color.  Gay employees of color were less likely than their White counterparts to disclose their sexual orientation at work, but lesbians were as likely as gay men to disclose at work.  Race and gender differences were not found for heterosexism, but supervisor and work group demography were related to disclosure and heterosexism.  Employees with heterosexual supervisors and co-workers reported more discrimination than those in gay teams, and employees with supervisors of the same race or sexual orientation were more likely to disclose than those with supervisors of a different race or sexual orientation.  More heterosexism was reported in environments involving male supervisors or work teams, and these effects were stronger for lesbians than gay men.  Irrespective of race, employees who worked in racially balanced teams reported less heterosexism than those in primarily White or non-White teams.  Gender similarity of supervisors or co-workers did not affect disclosure or discrimination, suggesting that gender similarity predictions grounded in relational demography theory may not apply to gay employees.

Heterosexism in the Workplace: Do Race and Gender Matter?

Although gay men and lesbians constitute between 4-17% of the workforce (Gonsiorek & Weinrich, 1991), we know very little about their workplace experiences.  In fact, even though gay employees constitute a larger proportion of the workforce than many other minority groups, sexual orientation has been excluded from most empirical research on workplace diversity (Badgett, 1996; Croteau, 1996).  Discrimination against employees who are gay, or simply appear to be gay, is legal in most workplaces (National Gay and Lesbian Task Force, 1996).  Without legal protection, gay employees are vulnerable to discrimination, and existing research indicates that between 25-66% of gay employees report sexual orientation discrimination at work (cf. review by Croteau, 1996).  However, these are probably conservative estimates because most gay employees do not fully disclose their sexual orientation at work for fear of discrimination (Badgett, 1996; Schneider, 1987). 

This situation may be worse for gay and lesbian employees of color.  These employees face discrimination not just because of their sexual orientation, but also because of their race, ethnicity and gender.  Diversity scholars observe that women of color may face a "double-jeopardy" in the workplace because of their ethnicity and gender (cf. review by Ferdman, 1999).  A key question is whether this becomes a "triple jeopardy" for lesbian women of color.  Existing theory and research provide little information on the combined effects of racism, sexism and heterosexism in the workplace.  Do lesbians of color face greater heterosexism than other gay groups because racism and sexism spills over into heterosexism, or is heterosexism relatively independent from these other forms of discrimination? 

In addition to discrimination, gender and race may also affect the decision to "be out," or the disclosure of a gay identity to others in the workplace.  Lesbians and gay people of color may be reticent to disclose their sexual orientation at work because of their fear of becoming susceptible to yet another form of workplace discrimination, and also because they are already highly visible because of their race and gender (Kanter, 1977).  This visibility may increase their chances of being targeted for heterosexism.  The disclosure of a gay identity at work is often done on a careful case-by-case basis; gay employees reveal their orientation in situations where they feel safe and to individuals whom they trust (Badgett, 1996; Friskopp & Silverstein, 1996).  Gay employees who are highly visible on the basis of their race and gender may fear that "coming out" to one coworker may result in a domino effect of coming out to the entire organization.  Do these factors combine to make lesbians and gay people of color less likely than their White gay male counterparts to disclose their sexual orientation at work?  The first purpose of this study was to answer these questions by exploring the effects of race and gender on reports of heterosexism in the workplace and the decision to disclose sexual orientation to others at work.

Although demographic variables of race and gender may impact workplace experiences, these effects do not occur in a vacuum, but are influenced by the broader context of work relationships.  In fact, a relational demography perspective holds that the individual's work experiences are shaped by the demographic composition of the manager-subordinate relationship and work team (Tsui, Egan & O’Reilly, 1992; Tusi & O’Reilly, 1989).  A number of observable and non-observable demographic characteristics have been investigated in studies of relational demography, including race, ethnicity, gender, education, age, attitudes and tenure (see review by Tsui & Gutek, 1999), but there has been no research on the effects of sexual orientation demography on work experiences.  Similarly, other scholars have identified work group composition as an important contextual variable to consider when examining the effects of multiple group memberships on interpersonal discrimination (Landrine, Klonoff, Alcaraz, Scott & Wilkins, 1995; Tsui & Gutek, 1999), but sexual orientation has been omitted from these discussions.  A second purpose of this study was to examine whether the race, gender, and sexual orientation of work groups and supervisors affect gay employees’ reports of discrimination and their decisions to disclose a gay identity at work.

Theoretical and Literature Review

Heterosexism, Racism and Sexism: The Effects of Multiple Group Memberships on Workplace Experiences

There is a lack of research on the effects of multiple group memberships on workplace discrimination.  The impact of race and gender are often explored in isolation, as if employees have either a race or a gender, but not both (Ferdman, 1999), and sexual orientation has been excluded from these discussions.  In the following sections, we first introduce the construct of heterosexism and provide a foundation for examining the relationship between heterosexism, racism and sexism.  We then review the existing theory and research on the general effects of multiple group memberships on interpersonal discrimination.  Following this, we examine how these relationships may transfer to work settings involving gay and lesbian employees.  We use these various perspectives to develop and test two competing models of the effects of multiple group membership on heterosexism and disclosure of sexual orientation in the workplace.  

Defining heterosexism and homophobia.  Although there are a number of different terms that are used to characterize antigay attitudes and discrimination (cf. Herek, 1984), two of the most common are heterosexism and homophobia.  Heterosexism is defined as “an ideological system that denies, denigrates, and stigmatizes any nonheterosexual form of behavior, relationship, or community” (Herek, 1990, p. 316).  Heterosexism incorporates anti-gay attitudes, prejudice and discriminatory behavior (Sears, 1997).

Homophobia is a popular term that is often used interchangeably with heterosexism.  Homophobia reflects the fear and aversion associated with homosexuality (Weinberg, 1972).  There is a lack of consensus on the relationship between homophobia and heterosexism.  Some authors view homophobia as the emotional component of heterosexism.  For example, Sears (1997) defines homophobia as “prejudice, discrimination, harassment or acts of violence against sexual minorities, including lesbians, gay men, bisexuals, and transgendered persons, evidenced in a deep-seated fear or hatred of those who love and sexually desire those of the same sex.” (p. 16).  Morin and Garfinkle (1978) view homophobia as involving the individual’s irrational fear, as well as a cultural belief system that supports negative stereotypes about gay people.

Other authors maintain that homophobia and heterosexism are independent constructs.  For example, Jung and Smith (1993, p. 14) observe that: “Although heterosexism is often accompanied by homophobia, no logical or necessary connection exists between the two.  People who are homophobic may not be heterosexist; those who are heterosexist may not be homophobic.” Jung and Smith also provide an analogy that is pivotal to this study: “Heterosexism is analogous to racism and sexism.  Homophobia finds appropriate analogies in racial bigotry and misogynism.” (p. 14) The debate as to whether homophobia is part of heterosexism may be similar to the debate as to whether racial bigotry and misogyny are part of racism and sexism.  It is clear that individuals can engage in racist, sexist or heterosexist behaviors for reasons other than fear; prejudice may be based on self-interests, beliefs, values, group norms or social institutions (Allport, 1954).  At issue is whether individuals can be fearful or even hate a group without being racist, sexist or heterosexist.  This discussion provides important insights into the similarities and differences between heterosexism, racism and sexism.

The relationship between heterosexism, racism and sexism.  A review of the literature reveals two perspectives on the relationship between heterosexism and other forms of social prejudice.  One perspective is that heterosexism shares a common root with racism and sexism (eg. Fernald, 1995).  This “common roots” perspective holds that heterosexism springs from the same social, cultural and political foundations as racism and sexism.  Specifically, racism, sexism and heterosexism are all forms of social prejudice, and all involve attitudes, beliefs and behaviors that result in the superiority of one group over another (Herek, 1990; Jones, 1972; Lott, 1995).

A second perspective is that heterosexism is different from racism and sexism.  There are at least three factors that support this “independence perspective.”  First, the affective component of heterosexism, homophobia, has no real counterpart in racism or sexism.  Homophobia is grounded in heterosexual’s fear that they are gay, may become gay, or may simply be perceived as being gay by others (Herek, 1984). The ability to conceal sexual orientation also fuels homophobia by allowing others to speculate about an individual’s sexual orientation.  This fear does not readily translate to emotions underlying racism and sexism; individuals usually are not afraid that they may become or be viewed as another race or gender.

Second, the stigma associated with homosexuality is different than the stigmas associated with other groups (Goffman, 1974).  The invisibility of sexual orientation may amplify "courtesy stigmas," which are stigmas received by associating with stigmatized groups (Goffman, 1974; Herek & Capitanio, 1996).  Heterosexuals who associate with gay employees may be assumed to be gay by others in the organization; this form of stigma by association does not occur on the basis of gender, and rarely occurs on the basis of race.  Along with the courtesy stigma, gay men also face an AIDS-related stigma.  Although education has dispelled many of the myths associated with HIV/AIDS, the fear of AIDS is intertwined with a fear of homosexuality in a way that has no real parallel for race and gender.

The third factor that supports the independence perspective is the condemnation of homosexuality by many religious groups.  Although various religious groups throughout history have promulgated racism and sexism, the current focus of religious-based heterosexism has no direct parallel with race and gender.  Jung and Smith (1993) observe that some religious groups view gay men and lesbians as unnatural or diseased and “proclaims them to be at the core of their very being abhorrent to God.” (p. 61).  A compounding factor is the idea that homosexuality is a “lifestyle choice” (Jung & Smith, 1993), and this concept of choice is an important predictor of anti-gay attitudes (Herek & Capitanio, 1995).  In contrast, race and gender are not viewed as immoral life choices that violate religious beliefs.

Herek (2000) identified two competing frameworks that have been used to understand gay prejudice, and which parallel the common roots and independence perspectives.  The “gay rights framework” holds that attitudes toward gay people are psychologically similar to attitudes towards other racial and ethnic minority groups.  Like other minority groups, attitudes towards gays are based on political and religious values, normative pressures from peers, and the degree of inter-group contact.  In contrast, the “gay liberation framework” holds that gay prejudice is less about attitudes towards others and more about attitudes towards one’s own sexual identity.  Under this framework, gay prejudice is due to the individual’s confusion about their own sexuality, and their fear of being labeled gay.  This internal anxiety becomes externalized in prejudice against gays.  Herek observed that both of these frameworks may be operative, and that anti-gay prejudice may be due to a common process underlying minority group attitudes, as well as an individual process that reflects fears and insecurities about one’s sexual identity.

Research on racist, sexist and heterosexist attitudes provides support for both the “common roots” and the “independence” perspectives.  In support of the “common roots” perspective, Henley and Pincus (1978) found significant relationships among attitudinal measures of racism, sexism and heterosexism.  In addition, Herek’s (1984) review of the literature indicates that while heterosexism shares many of the same predictors as racism and sexism (i.e., limited contact, conservative religious orientation, limited education), some predictors are unique to heterosexism (i.e., guilt about sexuality, permissiveness about sexuality, and prior homosexual behaviors).  Ficarrotto (1990) found support for both the “common roots” and “independence” perspectives in his study of 79 undergraduate students.  In support of the “common roots” perspective, he found a significant relationship between racist and sexist attitudes and an attitudinal measure of homophobia.  However, he also found support for the idea that negative attitudes toward gays reflects the unique dimension of sexual conservatism, or deep-seated, negative feelings about human sexuality.  It is clear that more research is needed to test these two perspectives.

A final area that can shed theoretical light on the relationships among heterosexism, racism, and sexism is the area of aversive/modern racism (Dovidio & Gaertner, 1986) and subtle/modern sexism (Benokraitis & Feagin, 1995).  In general, the modernist perspective on prejudice holds that since modern social values prohibit blatant expressions of racism and sexism, overt prejudice has gone underground and now surfaces in more covert ways.  For example, the theory of aversive racism holds that aversive racists express egalitarian values and truly believe they are not racist, but unconsciously harbor racist feelings that result in subtle, but potent, biases (Dovidio & Gaertner, 1986).  At the heart of aversive racism is the individual’s ambivalence over egalitarian beliefs regarding socially acceptable behavior on the one hand, and their negative feelings towards minority groups on the other.  These feelings are usually unconscious, and involve discomfort, disgust or even fear of the minority group.  Aversive racism may have a direct counterpart in aversive heterosexism (Winegarden, 1994).  In essence, aversive heterosexism holds that unconscious homophobia leads individuals to behave in heterosexist ways, even though they profess and believe that they are not heterosexists.

Although aversive racism and heterosexism may share some common psychological processes, a key difference between them is that the social norms prohibiting heterosexism are not as well established as norms prohibiting racism and sexism.  For example, it is still legal to discriminate on the basis of sexual orientation in most workplaces (National Gay and Lesbian Task Force, 1996), and six out of ten Americans believe that homosexuality is morally wrong (Newport, 1998).  In addition to differences in social norms, the emotional component of heterosexism may be stronger, more subconscious, and more personal than the feelings associated with racism and sexism (Herek, 1984).  One consequence of this may be that the ambivalence experienced by aversive racists may be less or different than the ambivalence experienced by aversive heterosexists. 

In sum, while racism, sexism and heterosexism may share common foundations in social prejudice, heterosexism may also be distinct from traditional and modern forms of prejudice.  The “common roots” and “independence” perspectives provides an important context for understanding the effects of multiple group membership on interpersonal discrimination, and is the foundation for our model on heterosexism in the workplace.  We now examine the effects of multiple group membership on interpersonal discrimination and heterosexism in the workplace.

Multiple group memberships and interpersonal discrimination.  While the relationships between heterosexism, racism and sexism provides one theoretical piece of the puzzle, we need to understand how an individual’s membership in multiple minority groups affects his or her experience of heterosexism in the workplace.  While his model does not address workplace discrimination, Ransford's (1980) theory of multiple group memberships provides a good basis for examining the effects of multiple group memberships on heterosexism in the workplace. 

Ransford (1980) proposed the "Multiple Jeopardy-Advantage" (MJA) hypothesis, which holds that members of multiple low-status groups (i.e., African-American females) may be faced with a double disadvantage, whereas members of multiple high-status groups (i.e., White males) enjoy a double advantage.  Ransford used this MJA hypothesis to present two competing scenarios of multiple group memberships.  The first is that the effects of membership in multiple low or high status groups are independent.  For example, he notes that women of color experience both racism and sexism, but these effects may be independent; women of color would therefore experience equivalent sexism as White women.  As applied to gay men and lesbians, we would expect that lesbians would experience equivalent heterosexism as gay men, and gay people of color would experience the same heterosexism as their majority gay counterparts.

In contrast, Ransford’s second scenario, "multiple jeopardy-advantage" (MJA), holds that group memberships interact synergistically so that the total impact of multiple group memberships is greater than the sum of individual memberships.  He proposes that: “Multiple jeopardy emphasizes that class, ethnicity, sex and age discrimination may combine to produce unique barriers that cannot be explained by any one of these hierarchies singly.” (Ransford, 1980, p. 280).  Under this scenario women of color would experience greater sexism than White women because of the synergistic effect of multiple group memberships.  As applied to the current study, lesbians and gay people of color could be expected to experience greater heterosexism at work than their majority gay counterparts.  Ransford (1980) also theorized that White males have a dual advantage position due to their race and gender.  In addition to main effects for race and gender, we would also expect an interaction resulting in White gay males experiencing less heterosexism than any other group.

There is little guidance from the literature as to which of these models best predicts interpersonal discrimination.  Landrine and her associates assessed the Multiple Jeopardy Advantage (MJA) hypothesis by reviewing interpersonal discrimination studies published in social psychology and feminist journals, by examining race and gender differences in 1991 Census salary reports, and by conducting a laboratory study on discriminatory reactions to television characters (Landrine et al., 1995).  Landrine found conflicting support for the interaction between multiple group memberships and concluded that one problem with the MJA hypothesis is that it assumes equivalency across low or high-status group memberships.  Her conclusion that the effects of multiple group memberships may be contingent on the type of membership is a key point to consider when examining sexual orientation as a group membership, and is congruent with our earlier discussion on the similarities and differences between heterosexism, racism and sexism.  We now revisit that earlier discussion and examine how the MJA hypothesis can be applied to heterosexism in the workplace.

Sexual orientation and multiple group memberships in the workplace.  There has been a lack of theory or research on the relationship between sexual orientation and other group memberships in the workplace.  Ransford's (1980) two competing scenarios of interpersonal discrimination provide a good foundation for examining this topic, and are congruent with the “common roots” and “independence” perspectives reviewed earlier.  We integrate these approaches and use them as a springboard to develop two competing models of the effects of multiple group membership on heterosexism and disclosure in the workplace.

We call the first model the"spillover model.”  This model holds that the effects of race and gender "spillover" into heterosexism in the workplace, and is in line with the MJA hypothesis prediction (Ransford, 1980) that membership in multiple groups compounds the advantages or disadvantages associated with individual group memberships.  This model is also aligned with the broader “common roots” perspective, which holds that heterosexism, racism and sexism are all forms of social prejudice that involve attitudes, beliefs and behaviors that result in the superiority of one group over another.  

A basic premise of the spillover model is that the common foundation of different forms of prejudice allows for the transference of discrimination from one form to another at work.  This is congruent with other perspectives on workplace diversity.  For example, women and employees of color are often excluded from social networks, and face increased visibility and negative performance attributions that increase their susceptibility to job discrimination (Kanter, 1977; Pettigrew & Martin, 1987).  These factors may also make gay employees of color reluctant to risk further discrimination by disclosing their gay identity at work, and prior experiences of racism or sexism may amplify this reluctance (Rosabal, 1996). 

Emerging norms on racism and sexism may also contribute to the spillover model.  Since legislation and societal norms prohibit blatant racism and sexism, these forms of discrimination may become channeled into more socially permissible forms of heterosexism.  For example, the use of a racial slur against a gay employee of color could result in immediate dismissal, but the use of a gay slur may not even warrant a reprimand in most workplaces.

A logical extension of the spillover model is that group membership not only penalizes employees who are members of low-status groups, but also helps those in high-status groups.  This view is aligned with Ransford’s (1980) proposal that upper class White males obtain multiple advantages from their high-status group memberships.  As applied to the workplace, the same perceptual and attributional processes that place women and employees of color at a disadvantage may give advantages to their majority counterparts.  White gay males, for example, may be assumed to be competent because of their race and gender.  Accordingly, although White gay males face heterosexism, the spillover model predicts that the status, power and privilege associated with their race and gender should buffer them from the full brunt of heterosexism, and decrease their experience of heterosexism in the workplace.  White gay males should therefore experience less heterosexism than any other gay group.

A competing model, which we label as the "independence model," holds that heterosexism is independent from other forms of workplace discrimination.  This perspective predicts that race or gender will not affect experiences of heterosexism or the decision to be out at work.  This model builds upon Ransford’s (1980) theory that group memberships are independent in their effects, and is also aligned with the view that heterosexism springs from a different foundation than racism or sexism.  As discussed earlier, there are a number of ways in which heterosexism is different from other forms of social prejudice.  First, heterosexism involves an emotional component that lacks a counterpart in racism and sexism.  Second, the concealability of homosexuality amplifies courtesy stigmas and may fuel homophobia.  Third, homosexuality is viewed as an immoral lifestyle choice that runs counter to many religious doctrines.  Finally, heterosexism may take a different form in the workplace than racism or sexism.  For example, the competence, intelligence, motivation or leadership of a female or minority employee is often automatically questioned as an expression of racism or sexism, but these questions are usually not raised in response to an employee’s sexual orientation.  In contrast, gay employees are often treated as an onerous oddity, and their identity becomes shaped by their sexual behavior.  The questions aimed at gay employees often focus on their sexual behaviors, whether they are HIV-positive, or whether they will become sexual predators at work (e.g., Friskopp & Silverstein, 1996; Woods, 1994), rather than questions about their job performance or competence.

These factors may combine to create a situation where the heterosexism faced by gay employees is not influenced by their other group memberships.  Under this model, lesbians would be as likely as gay men to experience heterosexism, and gay people of color would face equivalent heterosexism as their White counterparts.  In addition, White gay males would not gain a heterosexism buffer by their race or gender, and would therefore be as likely to experience heterosexism as their female and minority counterparts.  Similarly, the independence model would predict that since race and gender would not affect heterosexism, and heterosexism is related to disclosure decisions (Ragins & Cornwell, 2001), race and gender would therefore not affect the decision to disclose a gay identity at work.

Although both the spillover and the independence models are plausible, there is no research to support one perspective over another in predicting discrimination against gay employees.  We therefore assess the spillover model by testing the prediction that lesbians and gay people of color will be less likely to disclose their sexual orientation and will report more workplace discrimination on the basis of sexual orientation than their majority gay counterparts. 

We controlled for two important factors in our study.  First, since gay employees in organizations covered by legislation prohibiting sexual orientation discrimination are more likely to be out at work and report less workplace discrimination than those in organizations that are not governed by such legislation (Ragins & Cornwell, 2001), we controlled for this variable in our study.  It is also important to control for disclosure of sexual orientation when investigating reports of discrimination; a gay employee may be less likely to be the direct target of heterosexism if no one at work knows that she or he is gay.  Accordingly, we controlled for these variables in our test of the spillover model:

Hypothesis 1a:  Holding protective legislation constant, lesbians will be less likely to disclose their sexual orientation in the workplace than gay males, and gay people of color will be less likely to disclose their sexual orientation in the workplace than gay Caucasians.

Hypothesis 1b: Holding protective legislation constant, gay White males will be more likely to disclose their sexual orientation in the workplace than any other group.

Hypothesis 2a:  Holding protective legislation and disclosure of sexual orientation constant, lesbians will report more heterosexism in the workplace than gay males, and gay people of color will report more heterosexism than gay Caucasians.

Hypothesis 2b: Holding protective legislation and disclosure of sexual orientation constant, White gay males will be report less heterosexism in the workplace than any other group.

Workplace discrimination may be affected by more than an individual's group membership.  As we discuss next, the demographic composition of the work group is a critical variable that may affect reports of heterosexism and the decision to be out at work.

Relational Demography and Heterosexism in the Workplace

A factor that may be even more important than the gay employee's race or gender is the race, gender and sexual orientation of his or her supervisor and workgroup.  A relational demography perspective predicts that individuals who work with managers and teams that are similar to them will form closer work relationships than individuals who work in dissimilar work groups (Tsui, Egan & O'Reilly, 1992; Tsui & O'Reilly, 1989).  A number of observable and unobservable demographic characteristics have been investigated in studies of relational demography (e.g., race, ethnicity, gender, education, age, attitudes and tenure) (see: Riordan, 2001; Tusi & Gutek, 1999).  However, these studies have not examined sexual orientation as a demographic variable, and have assumed that the effects of other demographic variables generalize to gay employees.

An application of relational demography theory to sexual orientation in the workplace leads to the prediction that sexual orientation discrimination and the decision to disclose a gay identity at work may be affected by the relational demography of the work team and supervisory relationship.  Gay employees who have coworkers and supervisors of the same race, gender and sexual orientation should have a more supportive work environment than employees in work settings where they are the only people of their race, gender or sexual orientation.

Holding protective legislation constant, we therefore expect the following relationships:

Hypothesis 3:  Gay employees who share a similar sexual orientation, gender or race with coworkers will be more likely to disclose their gay identity and, holding disclosure constant, will report less workplace discrimination than employees who differ from their coworkers.

Hypothesis 4: Gay employees who share a similar sexual orientation, gender or race with their supervisor will be more likely to disclose their gay identity and, holding disclosure constant, will report less workplace discrimination than employees who differ from their supervisor.

Finally, we explore whether these relational demography effects are symmetrical for different groups.  Existing research suggests that the effects of demographic similarity may be different for majority and minority group members, and some research indicates that majority members have a more difficult time in diverse groups than minority members (see reviews by Riordan, 2001; Tsui & Gutek, 1999).  Riordan (2001) has called for more research that investigates nonsymmetrical demographic effects, but there is little theory or research to guide an investigation of these effects among gay employees.  In fact, asymmetrical demography effects may be quite different for gay employees.  For example, current research predicts that a heterosexual male may have a negative reaction to being in an all-female group (Riordan, 2001), but this reaction may be quite different in the case of a gay male in an all-female group.  Accordingly, we investigate this issue with a research question:

Research Question: Are relational demography effects among gay employees symmetrical for race and gender?

Method

Procedure and Respondents

Sampling procedure.  As part of a larger study on workplace diversity (Ragins & Cornwell, 2001), surveys were sent to a national random sample of 2,919 members of three national gay rights organizations in the United States.  Specifically, we sent surveys to 1,488 members of one of the largest gay civil rights organization in the nation.  In order to obtain a diverse sample, we sent an additional 681 surveys to members of a national gay Latino/a organization, and 750 surveys were sent to a national gay African-American organization.  A stratified random sampling technique was used in which equal numbers of men and women were selected by geographic area.  The surveys were mailed in 1997 and were completely anonymous; there was no identifying information on the surveys that would connect them to the respondents.  Two reminder postcards and a reminder letter were sent to all respondents.  A total of 334 surveys were returned unanswered for various reasons, the primary reason being undeliverable mail (283); 51 surveys were returned unanswered because respondents were retired, unemployed, self-employed, heterosexual or deceased.  Completed surveys were returned by 768 respondents, yielding a response rate of 30%.

Respondents.  Because this study investigated workplace discrimination against gay employees, surveys returned from the following groups were not used in the analyses: those indicating they were heterosexual (n=20) or unsure of their sexual orientation (n=3), those who were self-employed (n=99) or employed by a gay, lesbian, or bisexual organization (n=51), and those who were not employed in paid positions (i.e., retired, unemployed or volunteers) (n=61).  The final sample consisted of 534 respondents.

The sample consisted of 168 women and 363 men; 3 respondents did not report their gender.  The majority of the respondents considered themselves to be gay or lesbian (92.9%), as compared with bisexual (7.1%).  The racial and ethnic background of the respondents was 67.6% White (n=361), 15.2% Black (n=81), 12.2% Latino/Hispanic (n=65), .7% Asian (n=4), 1.1% Multiracial (n=6) and 1.1% other (n=6); 11 (2.1%) did not report their race¹.  The average age of the respondents was 41 years.  The respondents were highly educated, the majority held bachelor's degrees (38.6%), and many had master's (28.2%) and doctoral (17.9%) degrees.  Forty-one percent of the sample earned between $26,000 and $50,000 a year, and 24% earned between $51,000 and $75,000 a year.  The average tenure in their current organization was 9.3 years, and the average tenure in their current position was 6.1 years.  The majority of respondents held professional or technical jobs (68.5%) and managerial jobs (19.7%), the remainder of the sample were employed in clerical or sales positions (4.9%), service or craft (6.4%) or agricultural (.4%).  Respondents also came from a large range of industries, such as education (24.2%), health (17%), government (14.8%), service (12.2%), manufacturing (9.2%), finance/insurance (6.8%), arts/entertainment (4.8%), advertising/publishing (3.1%), travel (2.0%), human services (2.0%) and design/fashion (.9%).

Measures

The survey was developed and pre-tested on an opportunity sample of 28 gay and lesbian employees across the nation.  The pretest was used to ensure clarity, refine instruments, and select items.  The staff from the gay rights organizations that provided the mailing lists also critiqued and approved the pretest and final surveys.

Controls for protective legislation.  Respondents were asked the city and state in which they worked.  Because legislation varies by region, multiple sources were used to determine current legislation governing respondents' employers (Button, Rienzo & Wald, 1997; National Gay and Lesbian Task Force, 1996; Wald, Button & Rienzo, 1997).  The protective legislation variable was coded 0= not covered by protective legislation, or 1= covered by protective legislation; therefore higher values represent protective coverage.

Work group and supervisor demographic composition.  As recommended by Riordan and Shore (1997), we operationalized our demographic variables as the individual's demographic characteristic relative to his or her supervisor and work group.  A set of questions asked respondents to indicate whether their coworkers were mostly the same race or ethnicity as them (coded as 3), about equally balanced (coded as 2), or mostly a different race or ethnicity than them (coded as 1).  Parallel sets of questions and codings were used to assess the gender and sexual orientation of the work group.  Another set of questions asked respondent's if their supervisors were the same race or ethnicity as them (coded 1) or a different race or ethnicity as them (coded 0).  Parallel questions and coding were used to assess the supervisor's gender and sexual orientation.  Higher values therefore represent greater similarity of the respondent's race, gender and sexual orientation to his or her manager and work group.  Respondents were also given an option of "don't know" for reporting similarities of coworkers' and manager's race and sexual orientation.  These responses were re-coded as missing data in the analyses.

Workplace discrimination.  A review of the literature revealed no psychometrically established measures of sexual orientation discrimination at work.  A modified version of the Workplace Prejudice/Discrimination Inventory was therefore used to test the hypotheses and research questions (James, Lovato & Cropanzano, 1994).  The 15-item, single-factor instrument has established reliability and validity (James et al., 1994).  The inventory measures perceptions of race discrimination in the workplace, so items relating to race were replaced with items relating to sexual orientation.  Sample inventory items include: "Prejudice against gays and lesbians exists where I work,"  "At work I am treated poorly because of my sexual orientation,” and "Supervisors scrutinize the work of gay and lesbian employees more than the work of heterosexual employees."  Although the instrument measures both experienced and observed discrimination, prior studies have found that the instrument represents a single factor (James et al., 1994), and a principal components factor analysis conducted on the present sample also yielded a single factor with an eigenvalue of 7.49 accounting for 53.6% of the variance.  The instrument uses a 7-point Likert scale, with responses ranging from 7 (completely agree) to 1 (completely disagree).  Higher values therefore indicate greater reported workplace discrimination.  The coefficient alpha found in the present study for this instrument was .94.

Disclosure of sexual orientation at work.  Disclosure was measured with the following question: "At work, have you disclosed your sexual orientation to: (Please check one option): (1) no one  (2) some people (3) most people (4) everyone."  These four options were modified from similar "out at work" scales used by Croteau and Lark (1995), Levine and Leonard (1984) and Schneider (1987).  Higher values represent greater disclosure of sexual orientation at work.

Results

The correlations, means, and standard deviations for the variables are displayed in Table 1.  Our sample reflected substantial diversity in work group relationships.  For supervisor-subordinate relationships, 30% (n=159) of respondents reported that their supervisor was a different race or ethnicity than them, 68% (n=362) had supervisors of the same race or ethnicity as them, and 5 individuals did not know their supervisor's race.  In terms of gender, 43% (n=227) had a supervisor of the opposite sex and 57% (n=299) had a supervisor of the same sex.  In terms of the supervisor's sexual orientation, 85.8% (n=452) had heterosexual supervisors, 8.9% (n=47) had gay supervisors, and 5.3% (n=28) did not know their supervisor's sexual orientation.  With respect to work group demography, 24% (n=126) had coworkers that were mostly a different race or ethnicity than them, 18%  (n=97) worked in groups that were about equally balanced, 57% (n=305) had coworkers who were mostly the same race or ethnicity as them, and 1 respondent reported not knowing coworkers race or ethnicity.  For gender, 29% (n=154) worked in groups composed mostly of individuals of the opposite sex, 43% (n=224) were in gender balanced work groups, and 28% (n=148) worked in groups that were mostly the same gender as them.  For sexual orientation, the majority of our respondents  (89.3%, n = 474) reported that most of their coworkers were heterosexual, 6.6% (n=35) reported that their work groups were about equally balanced,  2.3% (n=12) reported that most of their coworkers were gay or lesbian, and 1.9% (n=10) did not know their coworkers’ sexual orientation.

There was also significant variation in the degree to which individuals were out at work: 11.7% reported that they were out to no one at work, 37% reported being out to some people, 24.6% reported being out to most people, and 26.7% reported being out to everyone at work. 

We also wanted to assess whether our African-American, Latino-American, Asian-American, and multiracial respondents differed in experienced discrimination and disclosure decisions.  Analyses of variance indicated that respondents of color did not significantly differ from one another in the dependent variables of disclosure of sexual orientation (F (3,152)=.552, ns.), or reports of sexual orientation discrimination  (F (3,150)=.796, ns.).  We therefore combined and re-coded the race variable as 1 (White respondent) or 0 (Respondent of color.) 

Hierarchical regression analyses were used to test the study's hypotheses.  Our first set of hypotheses tested the spillover perspective, which predicted that lesbians and gay people of color would be less likely to disclose their sexual orientation (Hypothesis 1a) and would report more heterosexism at work  (Hypothesis 2a) than their majority gay counterparts.  We also tested for a significant interaction of race and gender, resulting in White gay males reporting less heterosexism (Hypothesis 2b) and being more likely to disclose at work (Hypothesis 1b) than any other group.  As displayed in Table 2, only Hypothesis 1a was partially supported.  When holding protective legislation constant, gender was not significantly related to disclosure, but gay people of color were less likely to be out at work than their White counterparts.  No support was received for the other spillover hypotheses.  When holding protective legislation and disclosure constant, we found no significant race or gender effects in reports of heterosexism at work.  Moreover, the lack of a significant interaction between race and gender indicated that White gay males were as likely to disclose and reported equivalent heterosexism as other groups.

Hypotheses 3 and 4 examined the effects of the demographic composition of the work team and supervisory relationship on disclosure and reported discrimination.  Since women and people of color are more likely to be in the minority in work groups, we controlled for race and gender effects by entering the respondent’s race and gender in the second step of our hierarchical analyses.  As displayed in Table 2, we found partial support for Hypothesis 3.  Gay respondents who worked with mostly gay coworkers reported less heterosexism and were more likely to be out at work than respondents who worked in balanced or mostly heterosexual groups.  However, the gender or racial composition of the work team was not significantly related to discrimination or disclosure.  Relational demography had a somewhat stronger effect in the supervisor-subordinate relationship.  In partial support of Hypothesis 4, respondents with supervisors of the same sexual orientation or race were more likely to be out at work than respondents with supervisors of a different race or sexual orientation.  In addition, respondents with gay supervisors reported less workplace discrimination than those with heterosexual supervisors.  The gender of the supervisor did not affect disclosure or reported discrimination.  In short, we found significant demography effects for similarities based on sexual orientation and supervisor race, but no significant effects for similarities based on gender.

The finding that gay employees were more likely to disclose their sexual orientation when they had supervisors of the same race led us to ask a follow up question about whether this effect was independent of the supervisor’s sexual orientation.  In order to answer this question we entered the supervisor’s sexual orientation before the race similarity variable in a hierarchical regression analysis.  When controlling for the sexual orientation of the supervisor, along with the other control variables (protective legislation, respondent’s race and gender), the similarity in supervisor’s race variable maintained its significant relationship with disclosure of sexual orientation (Beta= .10, p<.05).  This indicated that irrespective of the supervisor’s sexual orientation and the employee’s race, gay employees with supervisors of the same race were more likely to be out at work than those with supervisors of a different race.

Our research question assessed whether these relational demography variables varied by the respondent’s race or gender².  To answer this question, we tested the interaction between all of the relational demography variables and the respondent’s race and gender in a series of regression analyses.  We entered the control variables in the first step of the analyses, followed by the main effect terms in the second step (respondent race, gender, and the relational demography of the team and supervisory relationship), and the interaction terms in the final step.  None of the interactions was significant for disclosure, indicating that the demography effects found for disclosure did not vary by the respondent’s race or gender. 

While race and gender main effects were not significant, we found three significant interactions for discrimination.  First, as indicated in Table 3, we found a significant interaction between the respondent's gender and the gender composition of the work team.  As displayed in Figure 1, both gay men and lesbians encountered the most heterosexism in work teams composed of primarily men, and this effect was amplified for lesbians.   Second, we found a significant interaction between the respondent’s and supervisor’s gender.  A plot of the adjusted means in Figure 2 revealed that both gay men and lesbians reported more heterosexism with male supervisors than female supervisors, and that lesbians with male supervisors reported significantly more heterosexism than any other group.  Combined, these two interactions indicate that heterosexism is more likely to be reported in work environments involving male supervisors and primarily male work teams, and that this effect is stronger for lesbians than gay men. 

The third significant interaction was between the respondent’s race and the racial composition of the work group.  As displayed in Figure 3, both White gays and gays of color reported the most heterosexism in primarily White work teams.  In addition, both White gays and gays of color reported less heterosexism in groups that were racially balanced, than in groups that were primarily White or composed primarily of people of color.  In order to ensure that these interactions were not due to the sexual orientation of coworkers or supervisors, we re-ran all the analyses controlling for this variable.  All three interactions retained significance.

Discussion

We used a national sample of gay and lesbian employees to examine the effects of race, gender and work group demography on reports of sexual orientation discrimination and disclosure of sexual orientation at work.  Our study breaks new ground in exploring the effects of multiple group memberships and relational demography on the workplace experiences of gay employees, and provides a foundation for future research and theory development on this under-studied population.

The Effects of Multiple Group Memberships on Heterosexism and Disclosure

We developed and tested two competing models of the effects of multiple group membership on reports of heterosexism and the decision to disclose a gay identity at work: the “independence model” and the “spillover model.”  The results of our study indicate more support for the independence than the spillover model.  In contrast to the spillover model, lesbians and gay people of color did not report more heterosexism than their White male counterparts, and lesbians were as likely as gay men to be out at work.  However, in support of the model, gay people of color disclosed their sexual orientation to fewer people at work than their White counterparts.  One explanation for this is that gay people of color are more likely to be in the numerical minority in organizations than women, and this visibility may increase the perceived risks associated with disclosure.  Gay employees of color may already feel that they are under a microscope at work because of their race, and may not want to feed the gossip mill by coming out at work.  They may fear that their visibility may promote a "domino-effect" in which their disclosure to a select group of individuals results in everyone in the organization knowing their sexual orientation.  Finally, gay people of color who encounter racism at work may fear that a revelation of a gay identity may “add fuel to the discriminatory fire."

The finding that lesbians and gay employees of color reported equivalent heterosexism at work as White gay males supports the independence model, which holds that the forms and functions of heterosexism make it distinct from racism or sexism.  A key distinction is that the invisibility of sexual orientation combines with sexual insecurities and identity conflicts to create an emotional reaction that has no real counterpart in race and gender.  Individuals are usually not afraid that they are or will become another race or gender, or that they will be perceived as being a different race or gender by mere association with a stigmatized group.  Another distinction is that unlike sexual orientation, race and gender are not viewed as immoral life-style choices that violate religious doctrines. 

While our study provides a good start, more research is needed that examines the underpinnings of racism, sexism and heterosexism in organizations.  Our study assessed race and gender effects, but future research could take the next step by comparing mean reports of sexism, racism, and heterosexism using a racially diverse sample of gay and lesbian employees.  We also need a more thorough understanding of the relationship between homophobia and heterosexism in organizations, and the antecedents and outcomes of these constructs. 

While memberships in multiple high-status groups may give some individuals an advantage, we found no evidence of this among the White gay men in our study, who reported equivalent heterosexism as other groups.  The race and gender of these men apparently did not buffer them from sexual orientation discrimination.  One reason for this may be that openly gay men may not be invited to join the “good old boys club.”  Moreover, the disclosure of a gay identity after entering this club may evoke heterosexist backlash; they may be viewed as “imposters who infiltrated the White heterosexual male bastion.”  In addition, our finding that White gay men were as likely to be out at work as their female and minority counterparts suggests that although coming out may mean relinquishing the privileges associated with their race and gender, the emotional costs involved with hiding a gay identity may make that decision well worthwhile.

Future research should use both quantitative and qualitative approaches to examine race and gender differences in identity management strategies.  Our study examined race and gender differences in the decision to disclose, but we did not explore the strategies used by those who concealed their sexual identity at work  (cf. Button, 2001).  While lesbians were as likely to be out at work as gay men, it would be interesting to examine the identity management strategies used by those that remained in the closet.  For example, Woods (1994) observes that two non-disclosure strategies involve avoiding the issue of sexuality and counterfeiting a heterosexual identity.  Are there race and gender differences in the use of these strategies?  One could argue that counterfeiting a heterosexual identity would be difficult for those who are already highly visible because of their minority status.  In addition, the number of minority group memberships may matter; individuals who are members of multiple stigmatized groups may be more adept at managing their identities than those who are members of just one stigmatized group. 

Future research could also explore the factors that predict disclosure in the workplace.  Lesbians were as likely as gay men to be out at work, but the factors that led to this decision may vary by race and gender.  We need to understand what factors build the sense of trust and safety necessary for lesbians and gay people of color to disclose at work.  Perhaps they disclose their sexual orientation to coworkers only after they have "tested the water" for racism and sexism.  These questions point to the importance of examining not just the race and gender of the gay employee, but also the demographic composition of their work environment.

The Impact of Relational Demography on Heterosexism and Disclosure

Although the race and gender of gay employees did not affect their reports of heterosexism, the demographic composition of their work environment emerged as an important predictor of disclosure and discrimination.   In fact, sexual orientation emerged as the leading demographic predictor of workplace experiences for gay employees.  In support of relational demography theory, gay employees with gay supervisors or primarily gay work groups were more likely to be out at work and reported less heterosexism than employees in heterosexual work environments. While sexual orientation is often omitted from discussions of relational demography, it is central to gay employees and should be included in future research.

Also in line with relational demography theory was the finding that gay employees were more likely to disclose their sexual orientation when they had supervisors of the same race or ethnicity, and this effect held regardless of the supervisor’s sexual orientation or the employee’s race.  This suggests that although gay employees of color may be more reluctant than White employees to disclose at work, having a same-race supervisor may lessen this reluctance.  Even if the supervisor is heterosexual, similarity in race may be sufficient to build the trust necessary for disclosure.  Gay employees may also view a supervisor of a different race as less motivated to protect them from the potentially negative consequences of disclosure. 

We found no support for relational demography predictions regarding gender similarity.  Gender similarity in supervisory relationships and work teams did not affect disclosure or reports of heterosexism.  On closer inspection, it becomes clear that the assumption of heterosexuality underlies relational demography predictions.  Gender similarity predictions are based on the idea that working with someone of the same gender increases an individual's sense of comfort, security and acceptance.  This comfort may not be afforded to gay employees; the experience of a gay male working in a group of heterosexual men may be quite different than the experience of a heterosexual male in a heterosexual male group. The prediction that gender similarity creates a positive group climate also assumes that individuals in same-gender groups do not encounter the potential for sexual tension that may be present in cross-gender groups.  This prediction is reversed for gay employees; a gay male working in group of gay males may encounter similar sexual tensions as a heterosexual male working in a group of heterosexual females.  This suggests that biological sex may be a poor predictor of perceived similarity among gay employees.  Moreover, biological sex may be an entirely meaningless construct for transgendered employees, whose gender identity is independent from their biological sex. 

Some of the most intriguing findings in our study involved the significant interactions betwee