Finding the right data
Since the mid-1990s, labour market discrimination based on sexual orientation has been investigated by a number of researchers (Badgett 1995, Black et al. 2003, Arabsheibani et al. 2005 and others). These studies frequently find sizeable earnings differences: 1) partnered gay men earn significantly less than partnered heterosexual men; and 2) partnered lesbian women earn significantly more than partnered heterosexual women.
Where individual level self-reports of sexual orientation are available, the earnings differences are generally smaller.
Due to limited data availability, the previous studies have struggled with a trade-off between representativeness and sample size. Couples-based data sets such as population censuses in Canada, the United States, and the United Kingdom yield very large samples of same-sex couples but do not identify the sexual identity of non-partnered individuals. In contrast, data sets with individual level information on sexual orientation or sexual behaviour have generally been much smaller in size.
The few studies with individual level information on sexual orientation and reasonably large samples of sexual minorities have been limited to single states (e.g., Carpenter 2005), limited to young adults (e.g., Plug and Berkhout 2004), or lacked information on labour market earnings (Carpenter 2008a). As a result, it has been difficult to know whether differences in estimated earnings effects of a minority sexual orientation in different studies are due to differences in the samples, populations, or outcomes. Relatedly, it has been difficult to disentangle alternative theories underlying sexual orientation-based differences in labour market outcomes (e.g., specialisation versus discrimination).
We overcome these challenges by using confidential versions of the 2012-2014 UK Integrated Household Surveys (IHS) to which high quality labour market earnings data from the country’s Annual Population Survey have been linked (Aksoy et al. 2017). To our knowledge it is the first countrywide dataset with both partnership status and self-identified sexual orientation combined with high-quality data on labour market earnings.
These data allow us to identify large samples of sexual minority individuals – over 2,500 self-identified lesbians, gay men, and bisexuals (LGB) – through responses to a direct question about sexual orientation. It also permits us to identify not only individual level sexual orientation but also same-sex partnerships. This means we can directly test for how measurement of sexual orientation (i.e., individual level self-reports versus same-sex partnerships) is related to earnings differences between sexual minorities and heterosexuals. These data also allow us to comment more directly on the possible explanations for earnings differentials.
Partnership status matters: full-time employment
We show that having data on both partnered and non-partnered sexual minorities is substantively important. Our full-time employment models indicate that: gay men are 4.5) percentage points less likely to be working full time than otherwise similar heterosexual men (while bisexual men are 11.9 percentage points less likely.)
Notably, this difference for gay men is driven by the partnered sample. Partnered gay men are 6.1 percentage points less likely to be working full time than otherwise similar partnered heterosexual men. In contrast, the difference for bisexual men is driven primarily in the non-partnered sample, where non-partnered bisexual men are 11.7 percentage points less likely to be working full time than otherwise similar non-partnered heterosexual men.
For women, we show that lesbians are 8.2 percentage points more likely to be working full-time than otherwise similar heterosexual women, while bisexual women are 5.4 percentage points less likely to be working full-time. As with gay males, the lesbian difference in full-time employment (although of opposite sign to that for gay males) is predominantly driven by the partnered sample. Partnered lesbians are 15.4 percentage points more likely to be working full time than similar partnered heterosexual women.
Partnership status matters: earnings
Turning to the earnings models, after controlling for observable determinants of earnings (such as education, location, and family structure), we find a positive and statistically significant earnings differential for partnered lesbians compared to partnered heterosexual women but no earnings differential for non-partnered lesbians compared with similarly situated non-partnered heterosexual women.
We find a negative and marginally significant earnings penalty for partnered gay men compared to partnered heterosexual men but no earnings differential for non-partnered gay men compared with similarly situated non-partnered heterosexual men. Taking together the overall population of both partnered and non-partnered individuals, we find that the earnings difference associated with a gay sexual orientation for men is near zero, while the associated population-based earnings difference among women associated with a lesbian orientation is a premium of about 5.5 per cent and is statistically significant.
Specialization or discrimination?
We argue that our results are consistent with specialization. Traditional heterosexual partnerships typically involve gendered specialization, with the man more engaged in market activities than the woman, particularly given the prevalence of children among heterosexual couples. Even if the degree of household specialization were the same in heterosexual and gay male households, gendered heterosexual specialization means that the average partnered heterosexual man will be more focused upon market activities than the average partnered gay man. By the same argument, the average partnered lesbian will be more focused upon market activities than the average partnered heterosexual woman. These differences should not accrue to non-partnered individuals. These specialization-based predictions hold in our data. Our findings that the lesbian premium among partnered individuals accrues approximately equally to lesbians who are household heads and lesbians who are not household heads also supports the idea that there is less specialization in a lesbian household.
While comparative specialization within the household is our preferred explanation for most of our results, there is some limited evidence for the presence of discrimination as an explanatory factor. Our results show that it is older gay men and partnered gay men that earn less than comparable heterosexual men. It is likely that the lack of a heterosexual marriage becomes more of a signal of sexual minority status as an individual gets older (Frank 2007).
Partnered gay men may also be more observable as being gay than non-partnered counterparts. They may have photos of a same-sex partner or list their same-sex partner as a beneficiary, for example. If there is discrimination against gay men, these more observable individuals may bear a greater penalty. Further, the gay male penalty only occurs outside London, where there is likely to be a stronger taste for discrimination. Finally, the bisexual male penalty only occurs in the private sector and not the public sector where there are greater protections against discrimination.
Taken together, then, our unique samples of partnered and non-partnered sexual minorities and high quality data on earnings provide novel evidence supporting a role for specialization in explaining sexual orientation-based differences in labour market earnings, with less evidence for selectivity and limited and mixed support for discrimination.
- This blog post is based on the authors’ paper Sexual orientation and earnings: new evidence from the UK, EBRD Working Paper No. 196, also forthcoming in Industrial and Labor Relations Review.
- For an EBRD podcast discussing the authors’ findings in this research, click here.
- The post gives the views of its author, not the position of LSE Business Review or the London School of Economics.
- Featured image credit: Gay, by Stanley Dai, under a CC0 licence
- Before commenting, please read our Comment Policy
Cevat Giray Aksoy is a Principal Economist in the Office of the Chief Economist at the European Bank for Reconstruction and Development in London and an Affiliated Researcher at the London School of Economics. His main research interests are inequalities in the labor market, female labor supply and economics of fertility. He tweets at @cevatgirayaksoy.
Christopher S. Carpenter is a Professor of Economics, Professor of Law, Professor of Education, Professor of Medicine, Health, and Society, and Professor of Health Policy at Vanderbilt University. He is an empirical labor and health economist by training and has published widely on the effects of public policies on health outcomes and on labor market outcomes for sexual minorities in the United States, Australia, Canada and United Kingdom. He is Associate Editor at Journal of the European Economic Association and the Journal of Health Economics and on the Editorial Boards of Journal of Policy Analysis and Management, American Journal of Health Economics, and Industrial Relations. He holds a PhD in economics from UC Berkeley and a BA in mathematics, economics, and public service from Albion College (MI). His research has been supported by the American Cancer Society, the National Institutes of Health, and the Robert Wood Johnson Foundation.
Jefferson Frank is a Professor of Economics at University of London, Royal Holloway College. Prof. Frank obtained his PhD from Yale and has previously taught at Essex University, Birkbeck College, Sussex University, Harvard University and the University of California (Berkeley and San Diego).