Abstract
How do lesbian, gay, bisexual, and transgender (LGBT) employees fare in US workplaces? Beyond formal discrimination, do LGBT workers encounter biases that degrade the quality of their day-to-day workplace experiences? Using a representative sample of more than 300,000 employees in 28 “best case” organizations—federal agencies with LGBT-inclusive policies—the authors examine not only whether these informal workplace inequalities occur but also where and for whom they are most exaggerated. LGBT employees report worse workplace experiences than their colleagues across 16 measures of employee treatment, workplace fairness, and job satisfaction. These inequalities are amplified or tempered by organizational contexts and can even affect turnover intentions. They are also intersectional: LGBT women and people of color have consistently more negative experiences than do men and white LGBT workers. These results help map the landscape of LGBT workplace inequality and underscore the importance of considering intersectional and organizational contexts therein.
Public opinion and state and local legislation regarding sexual identity and gender expression have changed dramatically over the past several decades (Sears and Mallory 2011). However, equality for lesbian, gay, bisexual, and transgender (LGBT) individuals in the United States has advanced unevenly. On the heels of the Supreme Court affirmation of same-gender marriage rights, scholars have turned their attention to other formal rights such as employment equality. Given formal discrimination in hiring and wages for LGBT workers documented in prior research (e.g., Badgett 1995; Badgett, Lau, Sears, and Ho 2007; Albelda, Badgett, Schneebaum, and Gates 2009; Tilcsik 2011), and given that fewer than half of US states have LGBT-inclusive anti-discrimination policies, such legislation would be a welcome protection for millions of workers.
Yet, as they do for women and racial/ethnic minority workers (e.g., Bobbit-Zeher 2011; Cortina et al. 2013), subtler interactional-level forms of LGBT inequality may pervade US workplaces. LGBT workforce inequality may manifest not only as formal discrimination in hiring, promotion, and wages documented in prior research (e.g., Badgett 1995; Tilcsik 2011; Anteby and Anderson 2014), but likely also as systemic informal degradations of the quality of LGBT employees’ day-to-day work experiences. Scholars understand little about these informal processes of LGBT workplace inequality (McFadden 2015). In this study, we ask, do informal workplace disadvantages accrue along LGBT status across multiple dimensions of workplace experiences? For whom, and in what contexts, are these inequalities intensified? We argue that LGBT identity, as a devalued status characteristic, can anchor negative beliefs about the competence and worthiness of LGBT individuals (Johnson, Markovsky, Lovaglia, and Heimer 1995; Ragins 2008). Such biases may mean that inequalities between LGBT and non-LGBT colleagues exist across an array of day-to-day workplace experiences and that these biases can be shaped by cultural and demographic factors in employing organizations.
Using representative data from a theoretically informative sector of the US labor force—federal employees—we examine workplace experience inequalities along three related but conceptually distinct dimensions: perceived treatment as employees—whether workers report respect and resources from supervisors and coworkers; workplace fairness—whether they perceive that their workplace operates meritocratically and without favoritism; and work satisfaction—whether they are satisfied with their jobs, coworkers, and work environments. Arguing that these LGBT inequalities are not identical across demographic categories or organizational contexts, we investigate the intersectional and contextual processes that influence where and for whom these inequalities are amplified.
Beyond these empirical findings, our study offers an opportunity to reflect more broadly on how workplace inequalities operate when devalued statuses are neither reliably visible nor behaviorally apparent. Unlike other status characteristics (e.g., gender, race, and age), LGBT status cannot always be read off of an individual’s physical presentation (Anteby and Anderson 2014). Yet, we argue that LGBT status may be linked to patterns of disadvantage similar to those of more visibly recognizable status characteristics. Additionally, LGBT status implicates processes of status management, or the management of whether and to whom one’s devalued status is disclosed (Johnson et al. 1995; Clair, Beatty, and MacLean 2005; Jones and King 2014), and underscores the potential burdens of such management even for identities that are not consistently visible.
We examine these processes of LGBT inequality in a “best case scenario” employment sector. Unlike the general US labor force, federal employees have been protected by anti-discrimination policies inclusive of sexual identity since 1998 and gender expression since 2012. Prior scholarship suggests that federal agencies’ formalized and bureaucratized accountability structures are generally more effective at undermining ascriptive bias than are less formalized hiring and evaluation procedures found in other sectors (Tomaskovic-Devey 1993; Bielby 2000; Stainback, Tomaskovic-Devey, and Skaggs 2010). As such, federal agencies may provide a conservative estimate of trends of LGBT status inequality in the US labor force overall. Given the uncertainty of LGBT-inclusive workplace protections within and outside the federal government, the results presented here may also understate informal workplace inequalities currently experienced by LGBT federal employees.
Through an analysis of 2015 survey data of a representative sample of more than 300,000 federal employees in 28 agencies, we identify widespread informal workplace experience inequalities: LGBT workers fare worse than do their non-LGBT colleagues in the same organizations on measures of perceived treatment, workplace fairness, and job satisfaction. Beyond analyzing broad LGBT workplace experience inequalities, we argue that these LGBT biases do not operate uniformly across the LGBT population or across organizations. 1 Past work demonstrates that the meanings attached to sexuality differ by racial identity (Pedulla 2014) and take on distinct characteristics in different occupational settings (Tilcsik, Anteby, and Knight 2015). Consistent with our predictions, we find that informal workplace inequalities are intersectional: LGBT status beliefs are racialized and gendered in ways that exaggerate these processes for LGBT-identifying women and people of color. This finding suggests the role intersectionality plays in LGBT status inequalities and underscores the importance of investigating ways LGBT status may moderate the gender and race workplace inequalities documented in prior scholarship.
Second, because the cultural and demographic makeup of organizations can influence the degree and types of stigmatization that occur within them (Bianchi, Kang, and Stewart 2012), we also argue that certain organizational contexts can amplify or temper LGBT workplace experience inequalities. For instance, organizational legacies of formal LGBT discrimination may foster workplace experience disadvantages, and demographic diversity within an organization may mitigate these workplace inequalities. Correspondingly, we find that, compared to other LGBT federal employees, LGBT employees in military-related agencies report significantly more negative workplace experiences, while those in agencies with the highest representation of LGBT workers have more positive workplace experiences. Counter to assumptions that diverse work environments “lift all boats,” however, greater gender and racial diversity alone does not mitigate LGBT inequality.
Third, we argue that these LGBT workplace experience inequalities may be significant enough to translate into long-term material consequences, in part by affecting LGBT employees’ career decision-making. Specifically, workplace experience disadvantages may contribute to LGBT employees’ consideration to leave their jobs in search of work in more welcoming environments. Indeed, we find that LGBT workers have higher turnover intentions than their colleagues and that workplace experience inequalities help explain this pattern.
Taken together, these findings advance scholarship on LGBT workforce inequality by documenting differences in informal workplace experiences by LGBT status across 28 separate organizations, and by examining the contextual and intersectional nature of these inequalities. The results also suggest ways that employing organizations, in addition to individual LGBT workers, may be negatively affected by these LGBT inequalities. LGBT employees’ more negative workplace experiences help explain their greater likelihood of intending to leave, and turnover is costly and disruptive for organizations (e.g., Moen, Kelly, and Hill 2011). LGBT individuals also report lower satisfaction compared to their colleagues, a measure often linked to productivity (Eisenberger et al. 2002). Finally, consistent with enduring patterns of gender and racial inequality (e.g., Cotter, Hermsen, Ovadia, and Vanneman 2001; Stainback and Tomaskovic-Devey 2009), our research foreshadows the persistence of disadvantages for LGBT employees even with a nationwide implementation of anti-discrimination legislation.
LGBT Status and Workforce Inequalities
We turn to literature on status inequality and status management to help theorize how LGBT identity may affect workplace experience inequalities. Status inequality theory asserts that interpersonal inequalities can emerge from various attributions of value and esteem based on broadly shared cultural beliefs about the “types” of people in different status categories (Ridgeway 2011, 2014). Similar to devalued race and gender categories, LGBT identity is a devalued status (Johnson et al. 1995) 2 that is accompanied by widely shared negative beliefs that range from heteronormativity, or normative assumptions about the naturalness of heterosexuality and sex binaries (Herek 2007); to stereotypes about LGBT individuals as aloof, untrustworthy, or incompetent (Cech and Waidzunas 2011; Dovidio and Fiske 2012); to more malevolent beliefs that LGBT individuals are irresponsible, lazy, deviant, and immoral (Herek 2007; Ragins 2008). More than half of Americans harbor some level of disapproval toward non-heterosexual sexuality (Smith and Son 2013; Doan, Loehr, and Miller 2014), and such beliefs are diffuse across a variety of social and institutional contexts (Ragins 2008).
As with other status characteristics, these broadly held cultural beliefs may bias how others assess the worthiness of their LGBT colleagues by shaping expectations and assessments of LGBT workers’ performance and by serving as the basis of exclusion of LGBT people from informal workplace social interactions (Johnson et al. 1995; Ridgeway 2011, 2014). These status biases, in turn, may mean that LGBT individuals are provided with fewer resources to do their jobs well and given less respect and mentorship from supervisors and coworkers. Status biases may also undermine LGBT workers’ assessment of the fairness of their workplaces and reduce their work satisfaction.
Because LGBT status is not always visible, LGBT-identifying individuals must often manage information about their status in social contexts (Badgett 1995; Ward and Winstanley 2005), deciding whether and to whom to disclose their LGBT status based on the perceived threats to disclosure (Clair et al. 2005; Jones and King 2014; Tilcsik et al. 2015). Because of near-constant presumptions of cisgender heterosexual identity, status management is an ongoing and often frustrating process that frequently re-occurs in new workplace interactions (Ward and Winstanley 2005). 3
LGBT workers may feel pressure to engage in status management tactics to “pass” (conceal) or “cover” (downplay) their LGBT identity in order to navigate potential workplace disadvantages (Waldo 1999; Ward and Winstanley 2005; Yoshino 2007; Schilt 2010). They may also simply prefer to hide their LGBT status (Schilt 2010). Nevertheless, these tactics do not make LGBT employees immune to informal interactional inequalities. For one, anxieties about the potential negative consequences of disclosure are accompanied by non-trivial emotional and cognitive costs (Clair et al. 2005; DeJordy 2008). Concealing one’s LGBT status also makes one more vulnerable to hearing heterosexist and transphobic comments from colleagues who assume they are in the company of heterosexual and/or cisgender others (Ragins 2008; Cech and Waidzunas 2011). Thus, many interactional-level processes of LGBT inequality may operate irrespective of whether one’s LGBT status is visible or salient to their coworkers and supervisors.
LGBT Inequality in the Workplace
Recent research has begun to document how LGBT inequality manifests in formal discriminatory benefits, hiring, promotion, and remuneration practices in US workplaces (e.g., Badgett 1995; Badgett et al. 2007; Albelda et al. 2009; Tilcsik 2011; Cech and Pham 2017). Less than half of US states prohibit workplace discrimination on the basis of sexual identity, and even fewer include protections for transgender workers (HRC 2017). Legislators in some states are currently working to walk back anti-discrimination protections passed at the local level. Additionally, the benefit policies in many US organizations exclude LGBT employees: One-third of Fortune 500 companies lack domestic partner benefits, and 72% do not have transgender-inclusive benefits (HRC 2015).
Discriminatory hiring and promotion practices based on LGBT status appear relatively common (Albelda et al. 2009; Tilcsik 2011; Mishel 2016). For example, audit studies have found that fictitious applicants who signal LGBT status are less likely to be invited to interview than otherwise identical non-LGBT applicants (Hebl, Foster, Mannix, and Dovidio 2002; Weichselbaumer 2003; Tilcsik 2011). More than 10% of LGBT survey respondents also report having been denied employment or a promotion because of their sexual identity or gender expression (LLP 2006; Badgett et al. 2007; Herek 2007). Further, same-sex cohabiting men experience a 9% salary penalty compared to other men (Badgett 1995; Arabsheibani, Marin, and Wadsworth 2007). Hiring and promotion discrimination are especially prevalent among transgender workers (Albelda et al. 2009), who face challenges finding and keeping employment (LLP 2006; Schilt 2010).
Beyond these formal discriminatory hiring and remuneration practices, informal, interactional-level biases likely also degrade the day-to-day workplace experiences of LGBT employees. For example, colleagues and supervisors may call LGBT employees’ competence into question, especially when stereotypes about LGBT individuals (e.g., gay men as effeminate) contradict masculine or feminine norms within occupational settings (see, e.g., Johnson et al. 1995; Cech and Pham 2017). LGBT employees may also face marginalization from casual workplace conversations and social events (e.g., Friskopp and Silverstein 1995; Ward and Windstanley 2005), isolating them from the networks and social capital that are vital for career advancement (Ragins and Cornwell 2001).
Although prior research has suggested particular ways that LGBT bias may affect work experiences, it is limited in its ability to speak to the prevalence and extent of these inequalities in the labor force. Because of the dearth of large, nationally representative surveys that include LGBT status, much previous research has relied on LGBT-only samples that do not allow for direct comparisons with non-LGBT workers (McFadden 2015). Even national surveys that do include LGBT identity measures can compare only samples of workers spread across the labor market, without the ability to understand LGBT employees’ experiences compared to their colleagues in the same organizations. Further, innovative audit studies (e.g., Hebl et al. 2002; Tilcsik 2011; Mishel 2016) have documented hiring discrimination, but these studies require by design that LGBT status be signaled in more obvious ways than typically occurs in workplace interactions.
Despite this pioneering past research, little knowledge is available about processes of informal LGBT workplace experience inequality. Earlier research suggests that LGBT employees tend to be less satisfied with their treatment and to feel marginalized (e.g., Firskopp and Silverstein 1995; Cech and Pham 2017; Lewis and Pitts 2017), but there is much left to be understood about these day-to-day informal inequalities and the contextual and demographic variability of LGBT inequality across organizations (Anteby and Anderson 2014; McFadden 2015). For instance, recent work comparing the experiences of LGBT and non-LGBT employees in STEM-related agencies in the federal government found that LGBT workers report more negative treatment and less workplace satisfaction than do their non-LGBT-identifying counterparts (Cech and Pham 2017). These findings bolster past work that documented persistence of heteronormativity and homophobia in STEM occupational contexts (Cech and Waidzunas 2011; Connell 2014). A question remains though: How do these processes operate in the workforce more broadly? Further, prior research has typically treated workplace heteronormativity and homophobia as uniformly consequential for LGBT workers across race and gender categories and organizational contexts. The present study aims to contribute to this understanding of LGBT workplace inequality by examining a variety of workplace experience dimensions across the entire sector of the federal government and scrutinizing intersectional and organizational circumstances that might affect where and for whom these inequalities are most prevalent.
Dimensions of Informal LGBT Inequality in Workplace Experiences
To help trace the contours of these informal workplace inequalities, we examine an array of measures across three broad dimensions of workplace experiences. These dimensions capture a range of workplace experiences shown to be important for the quality of workers’ day-to-day work lives. The first workplace experience dimension is respondents’ perceived treatment as employees (see, e.g., Masterson, Lewis, Goldman, and Taylor 2000; Eisenberger et al. 2002). As a result of their colleagues’ and supervisors’ negative status beliefs, LGBT employees may receive less respect from supervisors, less support and resources for their work, and less transparent performance evaluations (Lewis and Pitts 2017). They may also feel less satisfied with their pay, less supported in their attempt to balance work and family responsibilities, and less secure in their jobs to whistleblow (i.e., expose workplace misconduct to their superiors) than do their non-LGBT colleagues.
The second dimension is workplace fairness. Perceiving one’s workplace as fair is suggested to be correlated with important work outcomes, such as productivity and employee well-being (King and Cortina 2010). LGBT workers who encounter negative status beliefs may be less likely than their non-LGBT colleagues to see their workplaces as fair and meritocratic. Specifically, LGBT individuals may be more likely to feel as though their supervisors act with favoritism and less likely to believe that their workplace utilizes meritocratic advancement procedures.
The final workplace experience dimension is work satisfaction. Prior scholarship has connected poor work satisfaction to a variety of secondary outcomes including health, well-being, and turnover (e.g., Tett and Meyer 1993; Waldo 1999). Consistent with other research (e.g., Durst and DeSantis 1997), we not only attend to respondents’ personal satisfaction with their jobs but also to their satisfaction with workplace conditions, advancement procedures, and empowerment. LGBT employees’ greater likelihood of encountering negative biases, combined with the stress of status management, may mean that they feel less socially integrated and less personally connected to and satisfied with their work than do their non-LGBT colleagues. We therefore expect:
Intersectional Processes at Work
LGBT individuals are situated within a variety of intersecting identity categories (Collins 1990; Gamson and Moon 2004; Schilt 2010; Cech and Rothwell 2018). Because the cultural status beliefs of specific disadvantaged identities overlap (Ridgeway and Kricheli-Katz 2013), status beliefs about sexual minority and transgender individuals likely converge with gender and race biases (Gamson and Moon 2004; Cech and Waidzunas 2011; Moore and Stambolis-Ruhstorfer 2013). Although not everyone who falls within the same intersections will experience bias in the same way, it is important to consider how LGBT workplace experience inequality differs for LGBT individuals who occupy distinct locations along these intersections (Pedulla 2014).
The salience of gender and race inequalities in US workplaces likely means that LGBT workplace disadvantages vary across ascriptive categories. Women and racial/ethnic minorities face persistent biases in hiring, promotion, and pay (e.g., Cotter et al. 2001; Stainback and Tomaskovic-Devey 2009); are less likely to be seen as competent employees (e.g., Feagin and Sikes 1994; Correll, Benard, and Paik 2007); and are more likely to be denied respect and authority than are white men (e.g., Gorman and Kmec 2009). Women and racial/ethnic minorities are also more likely than white men to face marginalization from informal networks important for career advancement (Kanter 1977; Reskin and McBrier 2000; Attell, Brown, and Treiber 2017).
Emergent research suggests that LGBT workplace inequality may indeed intersect with gender and racial/ethnic inequality (Douglass and Steinberger 2015). For example, among transgender individuals, female to male transgender employees often report greater levels of respect and authority than those who transition from male to female (Schilt 2010). Additionally, recent work suggests that the effects of LGBT status may differ for racial minorities in unexpected ways. For example, Remedios, Chasteen, Rule, and Plaks (2011) and Pedulla (2014) found that gay black men may actually experience an employment advantage when compared to straight black men and gay white men—potentially because of the effect of intersecting stereotypes surrounding queer and black masculinity. It may be, then, that non-heterosexual men of color more closely align with notions of what it means to be an “ideal worker” than do straight black men or gay white men and correspondingly experience less status bias. Yet, these potential advantages may not translate into more positive workplace experiences for LGBT men of color, as “tokenized” groups in the workplace often have to manage their identities in specific ways so as to mitigate stigma and maximize potential advantages (Wingfield 2007, 2010; Dickens and Chavez 2018). Indeed, LGBT-identifying racial minorities are less likely to feel comfortable disclosing their sexual identity at work than are white sexual minority employees (Ragins, Cornwell, and Miller 2003).
Given the potential for negative status beliefs of racial/ethnic minorities and LGBT individuals to overlap to create additional binds for LGBT individuals of color (Gamson and Moon 2004; Moore and Stambolis-Ruhstorfer 2013; Ridgeway and Kricheli-Katz 2013), we expect:
Lesbian, bisexual, and transgender women likely may also encounter workplace experiences that differ from those encountered by gay, bisexual, and transgender men (Schilt 2010; Waite and Denier 2015). The strength and direction of those differences remain uncertain, however. On the one hand, lesbian women may have workplace advantages over heterosexual women and gay men (Blandford 2003; Baumle 2009; Waite and Denier 2015). On the other hand, as noted above, women face a variety of disadvantages in the workplace that most men do not. As such, we expect that LGBT-identifying women will generally, but perhaps not consistently, report more negative workplace experiences than LGBT-identifying men:
Following McCall’s (2005)“intragroup” approach to empirically investigating intersectional processes, we examine these racial and gender processes among the sample of LGBT respondents only.
Workplace Experience and Organizational Context
Scholars have speculated that workplace context may be an important factor in anti-LGBT discrimination (Ragins and Cornwell 2001), but little is known about how informal workplace inequalities depend on the cultural and demographic contexts of organizations. Organizations often have their own semi-autonomous culture (Kunda 2009); these cultures can foster unique constellations of status beliefs compared to what is common outside the organization (Bianchi et al. 2012). For example, specific organizations may be less tolerant of widespread status beliefs (e.g., heterosexist comments or jokes) than is typical among the labor force, or may accommodate or amplify status biases through their dominant norms and practices. The focal mission of the organization may also shape organizational cultural beliefs and practices in ways that temper or amplify common status beliefs (Cech and Pham 2017). Further, organizational practices and goals may select individuals with certain social beliefs into (or out of) the organization. These and other factors likely impact the types and severity of status biases within the organization (Kunda 2009).
Broadly, organizations whose cultural contexts offer more support for LGBT inclusion through demographic makeup, formal or informal policies, and interactional norms likely have fewer LGBT inequalities. Conversely, organizations whose cultural contexts buttress or re-inscribe negative LGBT status beliefs through their norms of conduct and informal practices likely have exaggerated patterns of inequality. Because our respondents are employed across 28 agencies, we are able to examine whether variation in a few particular organizational factors might affect the severity of LGBT bias therein. In particular, we explore two key contextual aspects that may affect LGBT status inequality—one that may amplify it and another that may temper it.
First, organizational contexts with legacies of formalized anti-LGBT discrimination may amplify the negative LGBT status beliefs of employees and supervisors. An ideal-typical example is military-related federal agencies in which the decades-old “Don’t Ask, Don’t Tell” (DADT) policy was only recently rescinded. Although the military has made progress in its treatment of sexual minorities and transgender individuals, the future of these rules is uncertain and little has been done to offset the lasting cultural effects of institutionalized discrimination (Burks 2011; Bowring and Brewis 2015). Consequently, this legacy of discrimination may mean that LGBT employees in military-related agencies face particularly poor informal work environments compared to LGBT employees in other agencies and non-LGBT individuals (Moradi 2009). Five federal agencies are categorized as military-related: the Departments of the Army, Air Force, Navy, Defense, and Homeland Security. 4 Active-duty military personnel were excluded from the survey. We therefore expect:
Second, greater diversity by sexual identity and gender expression may reduce LGBT status inequalities. The presence of similar others may help LGBT individuals find social support (Ragins and Cornwell 2001; Ragins 2008). Additionally, among heterosexual individuals, greater frequency of contact with sexual minorities tends to reduce sexual prejudice (Smith, Axelton, and Saucier 2009). LGBT-identifying employees in agencies with higher proportions of LGBT individuals may therefore be less likely to feel as though they are tokens of LGBT identity (Kanter 1977) and will find more LGBT colleagues with whom to network. Accordingly, we predict 5 :
Additionally, greater diversity along other identity axes—gender and race in particular—may similarly foster more positive experiences for LGBT individuals. Some literature on workplace diversity suggests that supervisors in organizations comprised primarily of white men tend to express more overt gender and race bias than do supervisors in workplaces with greater diversity (Blau 1977; Kanter 1977). As such, we expect:
Organizational context may alter the degree to which LGBT employees encounter status biases in their workplaces in many other ways. Hypotheses 4 through 6 thus test whether the level of bias experienced by respondents is sensitive to such contextual factors.
Workplace Experience and Turnover Intentions
Finally, we hypothesize that these inequalities in workplace experiences may affect the quality of LGBT employees’ work lives. Far from being simply an interactional nuisance, such workplace experience disadvantages may be consequential for LGBT workers’ careers by shaping the trajectories along which they steer their careers. For example, negative workplace experiences may contribute to LGBT employees’ intentions to leave their jobs. Turnover intentions, while not a perfect predictor, are highly correlated with actual turnover (Steel and Ovalle 1984). Experiences of (or fears of encountering) negative status beliefs may help explain why LGBT employees are more likely than their non-LGBT colleagues to seriously consider leaving their organization within the next year.
First, we expect that LGBT employees will, in general, have higher turnover intentions than their colleagues:
Furthermore, we expect that the workplace experience inequalities discussed above will help explain these higher turnover intentions among LGBT employees:
Significant outcomes would suggest that these workplace experience disadvantages have more long-term career consequences that can affect LGBT employees’ career paths in serious ways.
Method
Data
We use 2015 Federal Employee Viewpoint Survey (FEVS) data for our study. Although FEVS has limitations—namely, it is cross-sectional, it does not disaggregate specific LGBT subcategories, and does not offer measures of “outness” or more detailed job measures (Griffith and Hebl 2002; Smith and Ingram 2004; Schilt and Wiswall 2008)—to our knowledge, these data are the only available national-level, multi-organization data that allow for an investigation of LGBT workplace experience inequalities on the scale and to the depth we examine.
FEVS was completed by 392,752 workers employed in federal agencies that represent 97% of the executive branch of the US government (OPM 2015). The US Office of Personnel Management (henceforth, OPM) has administered the FEVS since 2002 and LGBT status was added to the FEVS in 2012. The 2015 FEVS was administered electronically to a representative sample of permanent, non-seasonal employees in 35 large agencies and 45 small/independent agencies. 6 The data include indicators for the 27 largest agencies and aggregate the remaining smaller agencies into an “other small agencies” category, for a total of 28 agency categories. The response rate was 46.8%, which is typical for workplace surveys (see, e.g., Baruch and Holtom 2008). The sample used for our analysis includes 330,414 respondents, 11,094 who identify as LGBT, which excludes those with missing data or a response of “prefer not to say” on the LGBT status question described below. 7
Per standard analytic procedures (Allison 2001), we use multiple imputation via Stata’s chained equations technique to handle missing data, thereby generating five multiply imputed data sets that we pool to produce the resulting coefficient estimates. Less than 10% of data on any given question was missing. 8 We weighted regression models with the proportional weight (“postwt”) provided by OPM.
Operationalization of Key Variable
Respondents were asked in the survey: “Do you consider yourself to be one or more of the following (mark all that apply),” with the following response options: “Heterosexual or Straight,”“Gay or Lesbian,”“Bisexual,”“Transgender,” and “Prefer not to say.” Respondents who selected gay, lesbian, bisexual, and/or transgender status were coded as “LGBT” by OPM. Respondents who selected heterosexual or straight and did not indicate transgender status were coded as “non-LGBT.” We excluded the 12% of respondents who reported “prefer not to say” from the main analysis but, as described below, we compare this group with LGBT and non-LGBT-identifying individuals in supplemental analysis. Consistent with 2013 Gallup poll estimates that 3.4% of the US population and 2.8% of those with college degrees are LGBT (Gates and Newport 2012), 2.97% of our sample identifies as LGBT.
Workplace Experience Measures
For the attitudinal questions, respondents were asked the extent to which they agree with each workplace experience statement (1–5 response range: “strongly disagree,”“disagree,”“neither agree nor disagree,”“agree,” and “strongly agree”). Satisfaction-related questions offered a parallel 1–5 scale ranging from “very dissatisfied” to “very satisfied.” To protect confidentiality, OPM re-coded response values for each question into a 1–3 positive/negative response range, where 1 = negative (strongly disagree or disagree; very dissatisfied or dissatisfied), 2 = neutral (neither agree nor disagree; neither satisfied nor unsatisfied), and 3 = positive (agree or strongly agree; satisfied or very satisfied). We analyzed the re-coded versions of the responses.
The 16 workplace experience measures used here include 11 computed scales and 5 single-question measures. Single-question measures are those that tap substantively important topics for which the survey does not have more than one question. There are 5 single-question, stand-alone items (see Table 1). The computed scales were created by first grouping topically similar questions into 11 substantively meaningful categories guided by existing literature on employee treatment, work fairness, and job satisfaction (e.g., Smith and Ingram 2004; Miceli et al. 2012; McFadden 2015; Cech and Pham 2017). The questions in each category were factor analyzed and questions that did not load onto a single factor or that differed substantively from the rest were removed. We validated categories with confirmatory factor analysis and discriminant validity tests. The remaining questions in each topical category were summed, and, to keep all scales within a 1 to 3 value range, we divided the sum by the number of questions included in the scale. Table 1 provides measure operationalization, question wording, and scale alphas.
Operationalization of Workplace Experience Measures
Notes: We conducted two additional scale validity tests beyond factor analyses. First, we used structural equation modeling to test for discriminant validity among scale measures in each workplace experience dimension; discriminant validity tests among the treatment, leader integrity, and work satisfaction measures were all significant at the p < .001 level. Second, using a bootstrap test, we factor analyzed and found alphas for two random 50% subsamples; none of the alphas between the two subsamples varied by more than 0.005 (or approximately 0.6%).
Organizational Context and Turnover Intentions Measures
The military-related agency indicator is a dichotomous measure, for which 1 = employed by Army, Air Force, Navy, Department of Defense, or Department of Homeland Security, or 0 = employed by another agency. FEVS does not include active-duty military personnel. We also created measures of the percentage of LGBT and the percentage of white men in each respondent’s agency. Specifically, we calculated the proportion of respondents in each agency who identify as LGBT and the proportion of respondents who identify as white men. From these proportions (hypothetically ranging from 0 to 100%), we created two new variables and assigned values to those variables for respondents according to the representations in their agency.
Turnover intentions is a dichotomous measure based on a question that asked respondents, “Are you considering leaving your organization within the next year?” (0 = no, 1 = yes).
Demographic and Work Measures
In addition to LGBT status, all models control for several demographic and work measures. Specifically, we control for gender, racial/ethnic minority status, and age cohort: gender: 0 = male, 1 = female; racial/ethnic minority status (re-coded by OPM in original survey data): 0 = non-minority (i.e., white), 1 = minority (i.e., Hispanic, Native American, Alaska Native, Native Hawaiian or Pacific Islander, Black, and/or Asian); age cohort (coded by OPM): 1 = under 40, 2 = 40–49, 3 = 50–59, 4 = 60 or above. Our models also control for all of the measures of job variation that were available in the FEVS data.
Consistent with prior research that suggests LGBT employees’ experiences can vary by job level and type (Smith and Ingram 2004; Schilt and Wiswall 2008), we control for tenure, supervisory status, and agency. We measure tenure using answers to the following question, “How long have you been with the federal government (excluding military service)”: 1 = 5 or fewer years, 2 = 6–14 years, 3 = 15 or more years; and supervisory status with, “What is your supervisory status”: 0 = Non-supervisor/Team Leader, 1 = Supervisor/Manager/Executive. Agencies in which respondents are employed are listed in Appendix Table A.1; agencies that had too few respondents to be listed separately were combined by OPM under the category “Aggregated Small Agencies;” a list of these small agencies can be found in the OPM report (2015).
Analytic Strategy
To test our first three hypotheses, we examine the difference between LGBT and non-LGBT employees’ responses on the three sets of workplace experience measures. Specifically, we run multilevel generalized linear models using the “gllamm” command in Stata 14, which can accommodate multiply imputed multilevel models with proportional survey weights. All models control for gender, racial/ethnic minority status, age cohort, tenure, supervisor status, and federal agency.
Following the intragroup approach to intersectionality discussed by McCall (2005), we tested the intersectional processes hypothesized in H2 and H3 by running models with racial/ethnic minority status, gender, and the other controls among the LGBT subsample only. To test for differences by organizational context (H4–H6), we ran multiply imputed ordinary least squares (OLS) and ordinal logistic models predicting the outcome measures with the following interaction terms: LGBT status × military-related agency indicator (H4), LGBT status × percentage LGBT in respondent’s agency (H5), and LGBT status × percentage white men in respondent’s agency (H6). The final analysis uses a multilevel “gllamm model” and generalized structural equation models (GSEM) to predict the effects of LGBT status and workplace experiences on turnover intentions (H7 and H8). To also estimate the indirect effects of LGBT status on turnover intentions through workplace experiences, we added a path from LGBT status to the workplace experience measure in each GSEM.
At the end of our results section, we describe several robustness checks we run to ensure our findings are robust, and we also address two key alternative explanations for our findings. Appendix Table A.3 summarizes our hypotheses and our empirical results.
Results
Table 2 provides means and standard deviations for our independent variables and controls. The 28 agency categories are listed in Appendix Table A.1, along with the proportion of the sample employed in each agency. LGBT-identifying employees make up 2.97% of the sample. This percentage is consistent with national statistics on the proportion of college-educated Americans (2.8%) who identify as LGBT (Gates and Newport 2012). A lower proportion of women and racial/ethnic minorities identify as LGBT than do the proportion of men and whites.
Means and Standard Deviations on Demographic and Turnover Measures for All Respondents, LGBT Respondents (N = 11,094), and Non-LGBT Respondents (N = 319,320)
The first hypothesis posits a direct relationship between LGBT status and measures of the three workplace experience dimensions. Table 3 presents multilevel OLS and ordinal logistic models predicting each workplace experience measure, grouped by workplace experience dimension. As predicted in H1, LGBT status is a significant and negative predictor of all measures related to treatment as employees (see models 1 through 7). Specifically, compared to their non-LGBT colleagues, LGBT employees report that their work success is fostered less often, they are less likely to report that they have transparent performance evaluations and adequate resources, and they feel less respected by their supervisors (significant at the p < .001 level). They are also less satisfied with their pay, less comfortable whistleblowing, and feel less supported in their attempt to balance work and life responsibilities (significant at the p < .001 level).
Regression Coefficients Predicting Treatment, Fairness, and Satisfaction Workplace Experience Measures with LGBT Status and Controls (N = 330,414)
Notes: Columns report unstandardized coefficients (and standard errors) from MLM gllamm regression models.
p < .05; **p < .01; ***p < .001 (two-tailed test).
Further, LGBT employees are less likely to report workplace fairness than are non-LGBT employees (see models 8–11 in Table 3). Net of demographic and job controls, LGBT respondents are less likely to feel that their work unit is meritocratic and supports diversity, they feel less positive about the integrity of their leaders, and they are more likely to see favoritism in operation than are their non-LGBT colleagues (significant at the p < .001 level). The remaining models (12–16) in Table 3 test whether LGBT employees have lower job satisfaction than non-LGBT colleagues. Supporting this expectation, LGBT employees are less satisfied with their work overall and with the working conditions, employee empowerment, and advancement procedures in their workplaces, compared to non-LGBT individuals (significant at the p < .001 level). 9
To place the LGBT status effect in context with other demographic differences in workplace experiences, we calculated Cohen’s d effect sizes (d = difference in means/pooled standard deviation) for the difference in means between LGBT and non-LGBT respondents, between women and men, and between racial minority and non-racial minority employees on each measure (see Appendix Table A.2). Suggesting the relevance of LGBT status alongside race and gender differences more typically examined in workplace inequality literature, the LGBT status effect sizes across the 16 workplace experience measures are comparable to—and often much larger than—the effects sizes for gender and racial minority status differences.
Our next set of hypotheses addresses intersectional race and gender differences. Table 4 replicates the models in Table 3 among the LGBT subsample only (N = 11,094), and we attend particularly to the effects of racial/ethnic minority status and gender. Consistent with H2, LGBT-identifying racial/ethnic minority respondents have significantly more negative workplace experiences than do white LGBT employees on 13 of the 16 measures. Specifically, compared with white LGBT respondents, LGBT employees of color perceive that their success and work-life balance is fostered less extensively, they have less transparent evaluations, they are respected less by supervisors, they are less likely to report that their workplace is meritocratic and their leaders act with integrity, and they report more negative experiences on four of the work satisfaction measures.
Regression Coefficients Predicting Treatment, Fairness, and Satisfaction Workplace Experience Measures among LGBT Respondents Only (N = 11,094)
Notes: Columns report unstandardized coefficients (and standard errors) from MLM gllamm regression models.
p < .05; **p < .01; ***p < .001 (two-tailed test).
Although the coefficients for gender are typically negative (Table 4), only six measures reach full statistical significance. Compared to LGBT men, LGBT women are less likely to report that they are respected by supervisors, less satisfied with employee empowerment in their workplaces, and less likely to report positive experiences on all four of the workplace fairness measures.
To visually represent these intersectional differences, Figure 1, panels A through C, present means on the three workplace experience dimensions for LGBT employees by racial/ethnic minority status and gender. The height of the columns represent, in order, the means for non-LGBT white men (for comparison), white LGBT employees, LGBT employees of color, LGBT-identifying men, and LGBT-identifying women. Asterisks indicate the significance of the female and racial minority status coefficients in the LGBT-only models as shown in Table 4 predicting each outcome, net of controls.
Workplace Experiences among White Non-LGBT Men and LGBT Employees by Gender and Racial/Ethnic Minority (REM) Status
To further explore whether three-way intersections by gender, racial/ethnic minority status, and LGBT status exist, we performed additional analyses. First, we sought to understand whether the race and gender patterns noted above are consistent among an additional axis of marginalization. As such, we reran the models in Table 4 among LGBT women only to see whether the main differences by racial/ethnic minority status remained significant. We find that racial/ethnic minority LGBT women reported significantly worse experiences than did white LGBT women on the same measures, for which we see significant racial/ethnic minority status differences in Table 4. We also find similar patterns of gender difference among racial/ethnic minority LGBT respondents except for two measures: among racial/ethnic minority LGBT respondents, there is no gender difference in whether respondents find personal satisfaction with their work, or whether they perceive that their organization is meritocratic.
Further, to understand whether LGBT women of color experience particular disadvantages beyond the race and gender differences documented in Table 4, we ran supplemental models among LGBT respondents with an interaction term between gender and racial/ethnic minority status. Here, we find two significant differences: LGBT women of color are significantly less likely than all other LGBT respondents to report satisfaction with pay (interaction term B = −.262, p = .042) but significantly more likely than other LGBT respondents to report willingness to disclose (B = .100, p = .006). We conclude that much more work is needed to disentangle these complex intersectional processes—particularly research that can disaggregate respondents by racial/ethnic category.
Workplace Experience and Organizational Context
Next, we examine the effects of organizational context on the workplace experiences of LGBT employees. Consistent with expectations (H4), LGBT employees in military-related agencies report significantly worse outcomes on several employee treatment measures (success fostered, transparent evaluations, respected by supervisors, and whistleblowing), two of the workplace fairness measures (diversity supported and leader integrity), and all of the workplace satisfaction measures (all significant at least at the p < .05 level); see column (1) in Table 5. Appendix Figure A.1 plots these interaction effects, one plot per model.
Regression Coefficients of Interaction Terms between LGBT Status and Military-Related Agency Indicator, %LGBT, and %White Men in Respondents’ Agency, for Models Predicting Treatment, Fairness, and Satisfaction Workplace Experience Measures (N = 330,414)
Notes: Columns report unstandardized (Unst.) coefficients and significance of interaction terms with LGBT status (column label) in ordinary least squares (OLS) and ologit regression models predicting each workplace experience measure (see row label). Models predicting each workplace experience measure were run separately and included controls for gender, racial/ethnic minority status, agency, supervisory status, employment tenure, and age category. OLS regression models were used to predict outcomes 1-4, 8-11, and 12-15; ordered logits were used to predict all other outcomes.
Agencies with the highest representation of LGBT employees: Department of Education (6.44%), U.S. Agency for International Development (6.47%), and the Department of State (5.35%). Agencies with the lowest representation of LGBT employees: U.S. Air Force (1.65%), U.S. Navy (1.8%), and the Nuclear Regulatory Commission (2.26%). In supplemental analyses excluding respondents in military-related agencies, the LGBT × %LGBT interaction terms showed the same patterns of significance as presented above, except that the LGBT coefficient for supervisor respect (4) and balance (7) are no longer significant.
Agencies with the highest proportion of white men: U.S. Air Force (50.81%), National Aeronautics and Space Administration (50.2%), and the Department of Transportation (49.3%). Agencies with lowest proportion of white men: Social Security Administration (21.02% white men), Department of Health and Human Services (21.63%), and the Department of Housing and Urban Development (22.84%).
We also ran these analyses separately for percentage women and percentage racial/ethnic minority in respondents’ agency. Although LGBT employees in agencies with the greatest representation of women fare better on several measures, LGBT employees in agencies with greater racial/ethnic diversity do not fare better than those in agencies with less racial/ethnic diversity.
p < .05; **p < .01; ***p < .001 (two-tailed test).
Further, we hypothesized that LGBT respondents in agencies with the greatest demographic diversity will have more positive workplace experiences (H5 and H6). Table 5 presents unstandardized coefficients and standard errors for the following interaction terms (which we ran in separate sets of models): LGBT status × the percentage of LGBT employees in respondent’s agency, and LGBT status × percentage white men in respondent’s agency. LGBT respondents in agencies with the highest representation of LGBT employees have significantly more positive workplace experiences than do others on nearly half (7 out of 16) of the measures (column (2) in Table 5). These interaction effects in each model are presented graphically in Appendix Figure A.2. By contrast, LGBT employees do not generally fare better in agencies with the lowest proportions of white men (column (3), Table 5).
Workplace Experience and Turnover Intentions
Our final two hypotheses address whether LGBT employees are more likely to consider leaving their organization than are their non-LGBT colleagues and whether the workplace experience inequalities documented above partially explain this response. Table 6 presents a multilevel gllamm model using LGBT status plus controls to predict whether respondents are considering leaving their organization in the next year (H7). As we expected, LGBT employees have significantly higher turnover intentions than their non-LGBT colleagues. Supplemental analyses (not shown) illustrate that, across all 16 workplace experience measures, the more positive respondents’ workplace experiences, the less likely they are to consider leaving their organization (all significant at the p < .001 level).
Multilevel Logit Model Predicting Turnover Intentions with LGBT Status and Controls (N = 330,414)
Note: Multilevel gllamm model.
p < .05; **p < .01; ***p < .001.
Additionally, using structural equation models (SEMs), we test for direct effects of LGBT status on turnover intentions and for indirect effects of LGBT status on turnover intentions through each of the workplace experience measures. We also run SEMs that include all the workplace experience measures in each dimension in the same model and, finally, a SEM that includes measures from all three dimensions in a single model. Table 7 provides the unstandardized coefficient estimates of the focal direct and indirect effects in these models. The significant indirect effects indicate that LGBT employees’ more negative workplace experiences partially explain why they are more likely to consider leaving their organization (H8). The LGBT effect is not fully mediated by workplace experience differentials, however, as the LGBT status direct effect remains significant in each model.
Multilevel Structural Equation Models Predicting Turnover Intentions with LGBT Status, Workplace Experience Measures, and Controls, with Parameter Estimates and Significance Levels for Direct and Indirect Effects
Notes: Models were run in the generalized structural equation model (GSEM) function in Stata 14, which allows for multilevel predictors. Models in this table replicate the model in Table 6 and add a regression path between LGBT status and the workplace experience measure in each model. Controls include agency, gender, racial/ethnic minority status, supervisory status, tenure, and age. Bootstrapping methods were used to produce bias-corrected indirect effects, standard errors, and p values.
p < .05; **p < .01; ***p < .001 (two-tailed test).
To summarize the results testing these hypotheses, Appendix Table A.3 lists each hypothesis and the corresponding empirical patterns demonstrated above.
Robustness Checks and Alternative Explanations
To check the robustness of these findings, we re-ran the analysis using five additional modeling strategies. First, instead of using multilevel models with individuals embedded in agencies, we ran the analyses with agency fixed effects with individual dichotomous controls for each agency. Second, instead of using multiple imputation as is recommended practice (Allison 2001), we ran each model with listwise deletion. Third, we randomly selected 10% of the sample and re-ran the analyses with only that 10% sample. Fourth, we re-ran the analyses with SEM, using latent measures for the workplace experience measures 1–7, 8–10, and 12–15 instead of factor-analyzed scales. Finally, we tested for possible cross-equation correlations using the Seemingly Unrelated Regression (SUR) approach. Specifically, we re-ran the models in Table 3 with the SUREG command with agency fixed effects (as gllamm does not accommodate the SUR option), running one SUREG for each of the three work experience dimensions: treatment, workplace fairness, and job satisfaction. Patterns presented above were reproduced using each of these five alternative modeling strategies, suggesting that the findings are robust to variation in analytic approach and sample size. (These additional models are available upon request.)
A possible alternative explanation of the LGBT status effects presented above could be that LGBT respondents have uniformly more negative attitudes about their work that are unrelated to workplace bias. To test this explanation, we conducted supplemental analyses with a set of questions that asks respondents who participate in five particular employee programs—alternative work schedules, health and wellness, elder care, child care, and employee counseling programs—how satisfied they are with these programs. If LGBT respondents tend to be more negative about their work experiences across the board, we would expect that LGBT status would also negatively predict satisfaction with these programs. Instead, in our additional analyses (which are available upon request), we find that LGBT status is unrelated to satisfaction with any of the programs. Unstandardized coefficients for LGBT status in multilevel models predicting satisfaction with the programs (controlling for gender, racial/ethnic minority status, supervisory status, tenure, age, and agency) are always nonsignificant as follows: predicting satisfaction with alternative work schedules program (N = 122,707; β:–.087, p = .321); predicting satisfaction with health and wellness program (N = 87,969; β:–.081, p = .288); predicting satisfaction with elder care program (N = 45,216; β:–.003, p = .970); predicting satisfaction with child care program (N = 8,868; β: .015, p = .948); and predicting satisfaction with employee counseling program (N = 6,037, β: .077, p = .763).
Another possible explanation is that LGBT individuals are over-represented in jobs with the poorest work conditions and least adequate resources (see Ueno, Peña-Talamantes, and Roach 2013; Tilcsik et al. 2015). Our supplemental analyses do not find evidence in support of that alternative interpretation. Running models among supervisors only, LGBT status remains a significant and negative predictor in all models except the model that predicts personal satisfaction from work. Further, if these effects were driven by LGBT employees having less desirable jobs, controlling for poor working conditions should substantially reduce the LGBT status effect. In supplemental models, we re-ran the analyses in Table 3 (except for the working conditions scale [model 13]), adding controls for the following two rough proxies of work conditions: “physical conditions allow employees to perform their jobs well” and “employees are protected from health and safety hazards on the job” (1 = negative, 2 = neutral, 3 = positive). Net of these measures, LGBT status is still highly significant in each model. These supplemental analyses suggest that the trends documented above are not simply a result of the types of work LGBT individuals tend to be employed in, but of their day-to-day workplace experiences. (These additional models are available upon request.)
Finally, we suspected that some respondents who identify as non-heterosexual or transgender may have felt more comfortable answering “prefer not to say” to the LGBT status survey question. As such, we anticipated that those respondents might also have more negative workplace experiences than those who identify as non-LGBT. In supplemental models, we re-ran the analyses in Table 3 and found that those who reported “prefer not to say” have significantly more negative outcomes on all 16 workplace experience measures (all significant at the p < .001 level) than do those who identify as non-LGBT, controlling for gender, racial/ethnic minority status, supervisory status, tenure, age, and agency. Consistent with literature that suggests people of color are less likely than whites to identify publicly as LGBT (Moore and Stambolis-Ruhstorfer 2013), we also find that racial/ethnic minority respondents have a higher representation among the “prefer not to say” category than the LGBT category. 10 Older and more tenured respondents are also more likely to report “prefer not to say” than to report their LGBT status, consistent with younger adults’ greater willingness to publicly identify as LGBT (Risman 2018). Although we cannot disentangle “prefer not to say” respondents who personally identify as LGBT from those straight cisgender respondents who refused to answer, we would argue that the most vulnerable sexual minority and transgender individuals, who in turn likely have more negative workplace experiences than do less vulnerable LGBT workers, are more likely to report “prefer not to say” than are those individuals who are less vulnerable. As such, the results from Table 3 may underestimate the strength of the actual LGBT status effect on workplace experiences.
Discussion
In our analyses, we find evidence of a variety of workplace experience inequalities for LGBT employees in federal agencies. Across 16 separate measures, LGBT employees report worse treatment and less respect, perceive less fairness in their workplaces, and have lower levels of job satisfaction than their non-LGBT colleagues, after controlling for gender, racial/ethnic minority status, supervisory status, tenure, age, and agency. These inequalities are not uniform across respondents or organizations, however. As hypothesized, LGBT workplace inequalities intersect with other status biases: Racial/ethnic minority LGBT employees have more negative workplace experiences than do white LGBT employees on nearly all measures. The consistency of these intersectional outcomes indicates important but under-researched processes whereby racial and LGBT biases overlap to amplify status disadvantages for LGBT individuals of color. Although several workplace experience measures also vary by gender, the pattern is less consistent. It is possible that LGBT-identifying women encounter some intersectional freedoms that provide opportunities for agentic action and counteract some of the well-documented processes of gender status inequality (Ridgeway and Kricheli-Katz 2013), at the same time that LGBT-identifying men do not enjoy the same gender privileges as heterosexual, cisgender men (Herek 2007). Much more work is needed to parse out these intersectional processes among LGBT employees. These results underscore that scholars cannot investigate the operation of LGBT workplace inequality without understanding it as a racialized and gendered phenomenon. Nor can the workplace inequality literature fully articulate disadvantages for women and racial/ethnic minority workers without understanding how these disadvantages are moderated by sexual identity and gender expression.
We argue that the prevalence of LGBT informal workplace inequalities depends in part on the cultural and demographic contexts of organizations. On the one hand, some organizational contexts may increase inequalities in workplace experiences, as we find in military-related agencies formerly under “Don’t Ask, Don’t Tell.” Legacies of heterosexist and transphobic policies and practices may linger in the culture of organizations, even when such policies have been formally revoked. On the other hand, some contexts may reduce these inequalities: We find that organizations with greater representation of LGBT employees have less extensive LGBT inequalities. This outcome could likely be the result of a co-constitutive process whereby greater LGBT representation fosters better workplace experiences which, in turn, leads to more effective recruitment and retention of LGBT-identifying employees. However, greater gender and racial/ethnic diversity does not appear to improve day-to-day experiences of LGBT employees. In a multitude of other ways, organizational context may shape the manifestations of LGBT inequality in a given workplace, including informal policies, the selection of workers into the organizations, and the norms and traditions around the organizations’ mission. Future research is needed to fully theorize these organizational context effects.
Finally, we find that LGBT workers are more likely to consider leaving their organizations than are their non-LGBT colleagues and that their more negative workplace experiences partially explain these turnover intentions. Although we do not know precisely why LGBT individuals intend to leave, the significant indirect effects of workplace experiences suggest that these informal LGBT workplace inequalities may contribute to more material disadvantages by encouraging LGBT individuals to seek employment elsewhere.
Beyond underscoring the importance of considering intersectionality and organizational context in LGBT workplace inequality literature, our findings emphasize the need to continue to deeply theorize the role of visibility of disadvantaged statuses within workplace inequality. It was an open empirical question whether LGBT status would anchor informal workplace inequalities in the same way as more visually apparent characteristics, such as gender and race. Theoretical conceptualizations and experimental tests of status inequality models have assumed that immediate categorization on the basis of visual or behavioral cues is important to the perpetuation of status inequalities (e.g., Ridgeway 2011, 2014). Our findings also suggest that devalued statuses need not be consistently visible to anchor informal workplace inequalities. This outcome hints that workplace experience disadvantages may accompany other devalued statuses that are frequently invisible, such as mental illness. Additionally, our findings suggest the need for more research on the interconnections between status management and interactional-level workplace inequalities (Clair et al. 2005; Reid 2015). Individuals’ management of their devalued status can create non-trivial levels of stress and anxiety and other personal and career consequences (DeJordy 2008; Jones and King 2014). Future research should seek to better theorize how these individual-level responses to status inequality may themselves be manifestations of that inequality.
Limitations
Although the data we analyzed in our study provide a new opportunity to understand LGBT workforce inequality among a representative sample of an entire sector of workers, they do have several limitations. First, as noted above, redactions by OPM to protect confidentiality means that the data cannot be disaggregated by LGBT category, nor by occupation, geographic region, or specific racial/ethnic category. We encourage further research to try to better understand these additional sources of variation as well as additional ways in which organizational and occupational contexts may ultimately shape the workplace experiences of LGBT employees.
Conclusion
Our study helps map the landscape of informal LGBT workplace experience inequalities—documenting not only whether they occur for LGBT federal employees across multiple dimensions of workplace experiences, as they typically do for more reliably visible statuses such as gender and race, but also where and for whom these inequalities are most prevalent. It also underscores the value of considering how inequalities in day-to-day workplace experiences, in addition to processes of formal discrimination, may disadvantage LGBT workers.
Beyond being an issue of inequality, these LGBT workplace experience disadvantages may have consequences for organizational effectiveness. We find LGBT employees are more likely than non-LGBT employees to consider leaving their organizations—a pattern partially explained by their more negative workplace experiences. Turnover is expensive: Organizations that lose talented employees due to negative workplace experiences not only perpetuate these forms of inequality, they also undermine their own economic competitiveness (Moen et al. 2011). Additionally, these LGBT status disadvantages may undercut organizational efficiency in other important ways. For example, organization scholars have shown clear linkages between employee satisfaction and worker productivity (Eisenberger et al. 2002). As such, workplace biases may mean that the talents of LGBT employees could go underutilized (Smith and Ingram 2004).
We also find that LGBT employees generally feel less comfortable whistleblowing than do their colleagues. LGBT employees’ greater discomfort in disclosing legal and regulatory violations may be particularly consequential for organizations, like many of these federal agencies, that are responsible for core aspects of public safety, health, and welfare (Miceli, Near, Rehg, and Van Scotter 2012).
Although organizational interventions are never without cost, individual organizations can do a number of things to help address these inequalities. Some organizations have initiated formal LGBT employee resource groups that advocate for LGBT-inclusive organizational policies and practices (Cech and Rothwell 2018), and others have instituted Safe Zone trainings that can mitigate anti-LGBT bias (Finkel, Storaasli, Bandele, and Schaefer 2003; Black, Fedewa, and Gonzalez 2012). Organization leaders themselves can work to confront negative biases that promote favoritism and unfair resource distribution. Finally, because prior research has shown LGBT bias to be even more pronounced in organizations that are not subject to LGBT anti-discrimination laws (Ragins et al. 2003), we propose that extending these formal protections to LGBT employees in all employment sectors should be a priority.
That we find these forms of workplace experience inequality among federal employees is important. Federal agencies set legal and cultural precedents for how diversity and inclusion are to be institutionalized in US workplaces and, unlike in other employment sectors, LGBT employees are protected under anti-discrimination legislation. This fact, in addition to the bureaucratized accountability structures of federal agencies, suggests that our results are likely a conservative estimate of LGBT workplace inequality in the US labor force overall, especially within sectors outside the purview of LGBT anti-discrimination legislation and those that typically have more informal accountability structures.
Ultimately, shifts in organizational practice, public policy, and popular attitudes are all required to reduce LGBT workforce inequality. Without addressing the more informal ways in which LGBT inequality is reproduced, US workplaces may fall short of the ideal of creating fair work environments for all employees.
Supplemental Material
ILRR_Cech_Rothwell_Supplemental-Online-Appendix – Supplemental material for LGBT Workplace Inequality in the Federal Workforce: Intersectional Processes, Organizational Contexts, and Turnover Considerations
Supplemental material, ILRR_Cech_Rothwell_Supplemental-Online-Appendix for LGBT Workplace Inequality in the Federal Workforce: Intersectional Processes, Organizational Contexts, and Turnover Considerations by Erin A. Cech and William R. Rothwell in ILR Review
Footnotes
Acknowledgements
We thank Marbella Eboni Allen, Jenifer Bratter, Sergio Chavez, John Cornwell, Justin Denney, Bridget Gorman, Elizabeth Long, Lindsey Trimble O’Connor, Robin Paige, Laura Rogers, Heidi Sherick, Tom Waidzunas, Alexander Watts, and Alison Wynn for useful feedback on previous drafts.
An Online Appendix is available at 
.
For information regarding the computer programs used for this study, please address correspondence to
1
Research in public administration (Lewis and Pitts 2017) has used federal employee data from 2012—before federal agencies had non-discrimination policies that fully protected LGBT employees—to explore LGBT employees’ perceptions of their treatment in federal agencies. Further, research using federal employees in STEM-related agencies as proxies for science and engineering occupations found that LGBT individuals experience more negative treatment and less positive work satisfaction than their colleagues (). We go beyond this important initial work theoretically and empirically by utilizing a more expansive set of workplace experiences and by focusing not only on whether LGBT employees across the federal labor force experience inequality relative to their non-LGBT peers, but for whom and in what contexts these inequalities are amplified.
2
Lesbian, gay, bisexual, and transgender identity categories may each act as their own status characteristics that are accompanied by certain status beliefs. We speak of “LGBT status” here because LGBT individuals are often aggregated into a single category in public opinion and discourse () and because of the aggregation of these identity categories in our data.
3
4
Although Veterans Affairs is military related, we exclude the VA because it was not formally covered by DADT and it primarily deals with health care and other benefits for veterans.
5
This relationship is likely bi-directional: Agencies with more positive workplace experiences for LGBT workers are likely also more successful at recruiting and retaining LGBT employees.
6
7
FEVS does not have formalized procedures for obtaining restricted-use data and OPM does not release these data because of concerns about the small sample sizes of bisexual and transgender individuals.
8
Given the small proportion of the sample that identifies as LGBT, we did not impute LGBT status for respondents with missing data on this question. Supplemental analysis that included imputed LGBT status produced substantively the same pattern of results.
9
Several of the racial minority status and gender indicators are significant and positive in , indicating that women and racial/ethnic minorities (REM) on average report more positive experiences on those measures. These are likely the result of the combination of occupational race and gender segregation and the better benefits and employee protections offered to non-professional workers in federal agencies than is typical in other sectors. FEVS redacts occupation, but when we restrict the occupational heterogeneity of the sample by looking only at supervisors, familiar patterns of inequality by race and gender emerge. Of the 16 measures, 11 are significant and negative for women (only 3 are positive) and 8 are significant and negative for REM status (only 2 remain positive). As we note below, however, the LGBT status effects in this supervisor-only sample are virtually identical to the main analysis with the entire sample. As such, we suspect that these positive effects by race and gender are more an artifact of the uneven distribution of women and racial/ethnic minorities into blue-collar and non-professional jobs in these agencies.
10
Although we suspect that the underreporting of LGBT status among racial/ethnic minority respondents actually underestimates the intersectional processes in , a counterargument could be that racial minority respondents who selected into the LGBT category have uniformly more negative attitudes about work and that this, in turn, artificially inflates the racial/ethnic minority effect. Our supplemental analyses suggest otherwise, however: Re-running the analyses on satisfaction with the five employee programs above with a Racial/Ethnic Minority × LGBT interaction term, we did not find that LGBT-identifying racial/ethnic minority respondents are any more critical of employee programs than are other respondents. This finding suggests that racial/ethnic minority respondents willing to report LGBT status (rather than “prefer not to say”) are not more likely to be “complainers” in ways that would inflate our intersectional results.
References
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.
