![]()
Part I
Wages and jobs
![]()
2
Discrimination based on sexual orientation
A review of the literature in economics and beyond
M.V. Lee Badgett
Over the last three decades a growing political movement among lesbian, gay, and bisexual (LGB) people has led to a heightened public debate about the existence of discrimination against LGB people and about the appropriate government response to actual or perceived discrimination.1 Academics in general, and economists in particular, have been slower to respond to the need to study discrimination. In the last ten years, however, economists have awakened to some of the interesting intellectual questions implicated by considering discrimination and sexuality. Recently economists have produced a flurry of empirical papers analyzing earnings differences by sexual orientation. Overall, the bulk of the evidence from studies by economists and others fits the hypothesis that lesbian, gay, and bisexual people face employment discrimination in the labor market in the United States and in some other countries.
Before reviewing the evidence gleaned from these studies, I will briefly review the historical and policy backgrounds and then outline some of the important methodological issues that shape the design and interpretation of specific studies, including measurement issues and possible endogeneity issues. Even though discrimination can occur in many different market and non-market contexts, almost all empirical research has been conducted with respect to the labor market, which will be the exclusive focus of this chapter.
History and policy context
At one time, asking whether lesbian, gay, or bisexual people were treated differently from similarly qualified heterosexual people in the labor market would have seemed a silly question, since many jobs explicitly barred gay people. Since some time after World War I, the United States has forbidden gay people to serve in the military (Eskridge and Hunter 1997). In the 1950s, government witch-hunts sought out and fired homosexuals in the State Department and in other security-related jobs (Johnson 1994â5). Openly gay people were (and still are in some places) banned from working with children, and such discrimination has been upheld by the judicial system (Eskridge and Hunter 1997:627â9). Licensing requirements for jobs as diverse as barbers or stockbrokers at one time barred individuals who engaged in activities seen as immoral or illegal, such as homosexual behavior (Teal 1971).
Some of those barriers to the employment of LGB people are now illegal in certain states and cities. Sixteen states (Hawaii, California, Nevada, Wisconsin, Minnesota, Illinois, Massachusetts, Vermont, New Hampshire, New Mexico, New York, Connecticut, New Jersey, Rhode Island, Maine, and Maryland) outlaw sexual orientation discrimination by private employers, as do 285 cities, counties, and government organizations (Human Rights Campaign 2003). One estimate is that roughly a quarter of the US population is covered by statewide sexual orientation nondiscrimination laws, and another 14 percent are covered by local laws (van der Meide 2000). Clintonâs 1998 Executive Order 11478 prohibited discrimination based on sexual orientation in civilian federal employment. Most of these new policies were enacted in the latter half of the 1990s, and the intensity of enforcement is not known, so their impact on the existence or degree of discrimination is also unknown.
Despite these legal changes, at least one form of differential compensation of LGB workers is still common and legal. Employers typically offer compensation that includes both wages and other noncash fringe benefits, such as family leave, vacation time, health insurance, and pensions. In 2000, for instance, 61 percent of employers offered health insurance benefits to full-time employees, and most of those workers can get health insurance coverage for their families (U.S. Bureau of Labor Statistics 2000). But since LGB employees cannot legally marry their same-sex partners, those partners are not eligible for available health insurance coverage or other marriage-linked benefits. This differential treatment of same-sex partners â even when those partnerships are virtually indistinguishable from marriage in terms of relationship commitment, longevity, and economic interdependence â constitutes one obvious form of explicit compensation discrimination against LGB employees.2 While a growing number of employers are beginning to offer spousal benefits to their employeesâ same-sex domestic partners, only 14 percent of firms that offer health insurance are willing to provide benefits to a same-sex domestic partner (Kaiser Family Foundation 2004), demonstrating that most LGB employees do not have access to partner benefits. This particular compensation discrimination issue has only recently been discussed at a public policy level and has not been the subject of much study by economists (see Badgett 2001 for a longer discussion of partner benefits).
In the current policy environment, the question of whether LGB people experience labor market-based discrimination has taken center stage in many debates about appropriate civil rights policies. Policymakers who see nondiscrimination laws as a burden to employers will not add protected categories lightly. Legislative hearings and lobbying highlight the issue of the existence of discrimination based on sexual orientation.
Methodological issues
Not all observers agree about the existence of pervasive wage or employment discrimination against LGB people, illustrating the need for further study by economists and others. Undertaking systematic studies that compare the workplace experiences of LGB and heterosexual workers for evidence of discrimination runs into several conceptual issues that complicate the development of valid and reliable survey instruments and sampling designs.
The first complication is defining what one means by âsexual orientation,â or being gay, lesbian, bisexual, or heterosexual. Sexuality encompasses several potentially distinct dimensions of human behavior, attraction, and personal identity, as decades of research on human sexuality have shown. Perhaps the findings from the 1992 National Health and Social Life Survey (NHSLS) reveal the complexity most clearly (Laumann et al. 1993). One group of respondents, 6.2 percent of men and 4.4 percent of women, report feeling sexual attraction to people of the same sex. A smaller group, 4.1 percent of women and 4.9 percent of men, have engaged in sexual behavior with someone of the same sex since the age of 18. An even smaller group, 2.8 percent of men and 1.4 percent of women, reported that they think of themselves as gay (or lesbian, for women) or bisexual. And the potential nesting is not necessarily complete or consistent; some people who have same-sex desires have never acted on them, and even a small number of men who think of themselves as gay or bisexual report no same-sex behavior or attraction, for instance.
For economists and other social scientists interested in survey-based comparisons of economic outcomes by sexual orientation, the different possible measures of sexual orientation obviously pose an empirical challenge. One approach is to choose the definition that best fits the social context being studied. For labor market interactions, self-identity might best capture a characteristic that could cause differential treatment by employers or fellow employees, since identity might influence labor market decisions and openness about oneâs sexuality in the workplace.3 For studies of health-related issues or studies of identity development, sexual behavior might be a better measure. Unfortunately, little actual choice exists for researchers since currently only the NHSLS offers a set of questions encompassing more than one dimension of sexuality for adults and including identity.
In practice, relatively few good datasets exist that are both probability samples and ask questions on sexuality. Researchers studying labor market issues who would prefer self-identity measures have used behavioral measures in two major ways. One approach that I have used with data from the General Social Survey (Badgett 1995, 2001) involves identifying LGB people based on the relative lifetime frequency of same-sex sexual behavior. In other words, I classify a person as LGB if he or she has had at least as many same-sex partners as different-sex partners since the age of 18. Other economists (e.g. Black and Makar et al. 2003; Blandford 2003) have used GSS data on sexual partners in the last year or last five years to classify sexual orientation, although such definitions have the disadvantage of not being able to classify people without recent sex partners.4 Blandford (2003) further separates âopenâ LGB people who are not currently married from the âmaskedâ LGB who are currently married.
While both methods are likely to sort individuals whose same-sex orientation is relatively weak into the âheterosexualâ category, the potential for misclassifying individuals is obvious but unavoidable, since the GSS provides no data on the length or nature of relationships. Therefore my measure probably misclassifies someone who had five very brief different-sex relationships while in college, followed by a 20-year same-sex relationship. The second measure probably misclassifies someone who ended a 20-year same-sex relationship and followed it with one brief different-sex relationship over the next five years.
The NHSLS, which did collect data on length of relationships, is too small to make definitive statements about which of these measurement strategies is more prone to misclassification. That survey suggests, however, that the first behavioral measure using behavior since age 18 comes fairly close to a measure of identity, since less than 16 percent of people with at least as many same-sex as different-sex partners considered themselves âheterosexualâ (Badgett...
