Thesis Title: Sexual Orientation Wage Gap in the United States of America Supervisor: prof. dr. Erik Plug Name: Artur Rymer Student number: 10579419 Track (within Economics and Business): Economics Field: Labour Economics
Sexual Orientation Wage Gap in the United States of America
Artur RymerThis document is written by Student Artur Rymer who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
I. INTRODUCTION
In the last three decades, the LGBTQ (lesbian, gay, bisexual, transgender and queer)
movement was able to introduce the discussion about the rights as well as discrimination of
sexual orientation and gender identity minorities to mainstream politics. Some of the
movement’s goals have been achieved in the area, including antidiscrimination laws in the public sphere and labour industry and samesex marriage in several countries. Particularly in the Western countries and to some extent in other parts of the world the pace and
magnitude of changes have been unprecedented, especially when compared to similar
examples in history such as the Civil Rights Movement or Suffragette movement. However, discrimination against the LGBTQ community still exists, even in countries in which rights
and protections have been extended to include and ensure equal treatment of sexual
orientation and gender identity minorities. On top of that, although much has been said and studied about the issue of LGBTQ rights and discrimination and the community’s place and situation within society, the focus has mostly been on the social, religious, legal and political
aspects, while the economic circumstances have been covered only be several research
papers so far. In particular, this results in our limited knowledge and understanding when it comes to earnings of LGBTQ people. As earnings directly influence one’s wellbeing and quality of life and are one of the most prominent indicators of a society’s and an individual’s
economic condition, studying them and identifying possible disparities is vital for an
assessment of welfare on a personal, social group and society levels. This is particularly
important for the LGBTQ community as, on the example of the United States, sexual
orientation minorities alone can constitute between about 2% and 14% of the population (Coffman, Coffman and Marzilli Ericson, 2013). If there exists discrimination in earnings in the labour market against LGBTQ people, then, with a social group of that size, not only individuals but society as a whole would be affected. For example, due to lower tax revenues or lower productivity resulting from lower wages. Therefore, studying the possible existence of an LGBTQ wage gap would allow to identify whether such discrimination indeed exists, what is its size and if and what kind of measures should be taken by the governments and the private sector to ensure the best possible social and economic outcomes. Moreover, it would provide further insights into the topic and allow for further research in the area of labour economics related to LGBTQ minority workers.
One of the reasons why the possible existence of an LGBTQ wage gap has so far not been extensively studied, despite the rising awareness of issues facing the community, is the difficulty of finding or constructing datasets that identify a person’s sexual orientation or gender identity (Laurent, Mihoubi, 2012). This is due to the fact that surveys in which respondents selfidentify their sexual orientation and gender identity are susceptible to bias
and misreporting (which can have particularly severe consequences for reliability of
research as LGBTQ populations tend to be small across samples), while data in which
sexual orientation and gender identity are identified through indirect inference (e.g. through partnership with a person of the same sex or through sexual relations with a person of the same sex) can potentially be imprecise. Additionally, as Plug and Berkhout (2004) argue, frameworks used in studies of racial and gender discrimination might be inappropriate for
studying the sexual orientation wage gap because, unlike race and gender, sexual
dataset it is difficult to distinguish between respondents whose employers know about their sexual orientation and those who do not. On top of that, the LGBTQ community is very
diverse and, therefore, there might exist considerable differences in earnings between
different groups within it. As a result, studying these groups separately is warranted, however, the problem with this is that the populations of interest within a sample could be too small for a reliable study. This paper will focus specifically on comparing earnings of gay men and lesbian women with their heterosexual counterparts, due to the fact that sexual
orientation minorities constitute the biggest share of the LGBTQ community (Coffman,
Coffman and Marzilli Ericson, 2013).
Furthermore, the United States of America is a country with a large population that is diverse in terms of ethnicity, age, level of education, affluence and, to some extent, regional differences. It is there that the seminal study on the topic of the sexual orientation wage gap
has been conducted by Badgett (1995) and, also, the US is the country where the
considerable share of research on the topic of the LGBTQ people in general and the
community’s economic situation in particular has been conducted. Among others, Antecol,
Jong and Steinberger (2007), Badgett (1995), Blandford (2003) and Coffman, Coffman and
Marzilli Ericson (2013). Therefore, focusing on the US allows for comparison of results between multiple studies in terms of time difference, magnitude and statistical significance,
especially in the context of rapidly changing attitudes towards the LGBTQ community.
Additionally, it provides the opportunity to find the most appropriate, precise and effective research methods for the topic at hand. On top of that, the US Bureau of Labor Statistics annually publishes American Time Use Survey that contains respondents’ key labour market indicators and personal information, thus allowing for easily accessible and reliable dataset, necessary for studying the sexual orientation wage gap at any point or across time. For these reasons, the US has been chosen as the country of interest for the purpose of this study.
Moreover, the aim of this study is to contribute to the existing research in three ways. Firstly, by investigating whether there exists a sexual orientation wage gap with the use of more recent data. This will permit replication of previous research and add further insights into the discussion. Secondly, by following Plug and Berkhout (2004) and conducting a broader analysis, in which earnings are compared across genders and sexual orientations simultaneously. This approach allows for a more insightful inspection of the problem at hand, by recognizing that there might be differences not only between heterosexuals and homosexuals but also between different sexual orientations and genders. Thirdly, by analysing different measures of earnings, namely hourly and monthly wages and annual household income, this paper attempts to find if, on top of differences in earnings, there exist differences in labour supply between different sexual orientations. That is, whether homosexuals can expect different earnings also due to different amount of hours worked.
Finally, the results of this paper, based on American Time Use Survey 2015, are as follows. Firstly, gay men can expect lower hourly and monthly wages than heterosexual men and these differences are statistically significant. Secondly, lesbian women’s hourly and monthly wages are not significantly different from heterosexual women and all women can expect significantly lower wages than heterosexual men. Thirdly, the differences in earnings between heterosexual men, gay men and women in general become larger in monthly
wages, compared to hourly wages, suggesting differences in labour supply across these groups. Finally, the results of the analyses of annual household income show that both gay and lesbian households earn significantly less than heterosexual households and that this income penalty is larger for lesbian households. This further adds to the possibility of differences in labour supply. These results are mostly consistent with the previous research on the topic of sexual orientation wage gap.
To summarise, this paper investigates whether there exist differences in earnings between
homosexual men and women and their heterosexual counterparts in the United States of
America.
The structure of the paper continues as follows. Firstly, an extensive literature review is presented in section II. Included in it are several studies on demographics of the LGBTQ community in the US and past research on sexual orientation wage gap, from both the US and outside the US. Secondly, the dataset, the variables used in this paper and the methods employed for extraction of these variables are explained in section III. Thirdly, the research methods used for the analysis of the sexual orientation wage gap are explained in detail and in order in section IV. Furthermore, the results of the analysis are presented in section V, followed by the discussion of the results in section VI, including the caveats of this paper and suggestions for further research. Finally, this paper is concluded with a summary of its findings in section VII.
II. LITERATURE REVIEW
For the same reasons that are behind the difficulties with conducting research on the sexual orientation wage gap, estimating the size of the LGBTQ population in general and gay and lesbian populations in particular can be problematic. This also applies to analyses of
demographics of these groups. Nonetheless, past estimates based on surveys in which
respondents were asked if they identify as lesbian, gay or bisexual, range from 1.7% to 5.7% of the population, while estimates for samesex attraction and samesex behaviour are 11%
and 8.8% respectively (Coffman, Coffman and Marzilli Ericson, 2013). However, as
mentioned above, a problem that arises with data collection in surveys with selfidentification of sexual orientation is potential underreporting due to respondents’ possible unwillingness
to provide information about their sexual orientation. This bias can be mitigated by
anonymity of surveys, however, Coffman, Coffman and Marzilli Ericson (2013) have
additionally used the ‘item count technique’ aimed at increasing accuracy of the estimates and compared the results with those of a usual direct questioning method. They have found that when when asked directly, 11.3% of people identified as nonheterosexual, 17.2% have
reported that they have had samesex sexual experience and 13.9% reported samesex
attraction. However, when applying the ‘item count technique’, the estimates for identifying
as nonheterosexual and samesex sexual experience were significantly higher at 18.6%
and 27.3% respectively, suggesting underreporting of sexual orientation and experiences
when applying traditional methods of data collection. Here it is important to note that samesex sexual experiences are not synonymous with homosexuality or bisexuality, proven by the fact that samesex attraction was not significantly higher with the ‘item count technique’. On top of that, the authors acknowledged that data used was not representative
of the US population as a whole, with young and liberal respondents particularly numerous in the sample (Coffman, Coffman and Marzilli Ericson, 2013). Nonetheless, based on the existing research, the size of the sexual orientation minorities as part of the population can be assumed to be in the range of 2% to 14%. Furthermore, when studying earnings, level of education is an important variable as the amount of years of schooling is positively associated with the size of income. Black, Gates, Sanders and Taylor (2000) have found that in the US both gay men and lesbian women are considerably more educated than their heterosexual counterparts. Based on 1990 US Census data, among gay men, 23.7% have a college degree and 13% have a postcollege degree. The estimates for married heterosexual
men are 17% and 10.3% respectively. For lesbian women these numbers are 25% and
13.9%, while for heterosexual women 16% and 6.1% respectively. Based on data from the General Social Survey and the National and Social Life Survey, the fractions of gay men and
lesbian women who have college or postcollege degree are larger than for their
heterosexual counterparts and this trend is consistent across all studied age groups. Higher education levels among homosexuals have also been shown in more recent studies such as
Denney, Gorman and Barrera (2013) and Liu, Reczek and Brown (2013). Therefore, if
education was the only determinant of income, gay men and lesbian women should be
considerably more affluent than heterosexual men and women.
With her seminal paper, Badgett (1995) has begun the academic discussion on the sexual orientation wage gap. In her work, based on the US General National Survey (GNS) from years 19891991, she has found that annually gay and bisexual men earned between 11%
and 27% less than heterosexual men and lesbian and bisexual women earned between
12% and 30% less than heterosexual women. However, although the result for gay and
bisexual men was significant at 5% level, the result for lesbian and bisexual women was not statistically significant. On top of that, Badgett (1995) identified homosexuals and bisexuals based on whether they have ever had a sexual partner of the same sex past the age of 18. This approach could cause imprecisions as some people who are in fact, heterosexual may be labeled as homosexual or bisexual and vice versa. Still, the then widely accepted assumption that homosexuals are more privileged and better educated and thus earn more than heterosexuals in the US has been proven to have no scientific grounds (Badgett, 1995). A more detailed study was conducted by Blandford (2003). Based on GNS data collected in years 19891996, with homosexuality and bisexuality identified through recent (less than 12 months) relationship with a person of the same sex, it tested for the existence of the sexual orientation wage gap. It also tested whether the gap can be partially attributed to differences arising from marital status (particularly because, at the time data was collected, samesex marriage was not legal in the US) and from the fact the fact that homosexuals, by definition, do not conform with the traditional gender norms. That is, whether the sexual orientation
wage gap can be explained by genderrelated expectations and differences in behaviour,
occupation and marriage. This hypothesis has been partially proven as Blandford (2003)
compared annual incomes and found that gay and bisexual men can expect about 30% lower earnings than married heterosexual men, while lesbian and bisexual women can earn
between 17% and 38% more than married heterosexual women but still less than
heterosexual men, with the results being statistically significant. Moreover, Allegretto and Arthur (2001) have studied the sexual orientation hourly wage gap, specifically in the context of differences between gay men and heterosexual men arising from marital status. Research
was based on 1990 US Census data and gay men in the sample were identified through relationship with a male partner. According to the results, gay men earn between 2.4%
(when compared with unmarried individuals) and 15.6% (when compared with married
individuals) less than heterosexual men and both bounds are statistically significant
(Allegretto and Arthur, 2001). However, when comparing earnings of married and unmarried
men in the whole sample, Allegretto and Arthur (2001) found a marriage premium of 14.1% for men in general. Therefore, the unexplained wage gaps between gay men and unmarried and married heterosexual men in the paper are 2.4% and 1.5% respectively. As a result, Allegretto and Arthur (2001) pointed out that the significant majority of the sexual orientation
wage gap for gay men is explained by marriage premium and the remaining unexplained
gap is relatively small. Furthermore, Antecol, Jong and Steinberger (2007) conducted a
study, in which, based on 2000 US Census data, they tested whether the sexual orientation wage gap can be explained by differences in occupation participation (occupational sorting), education (human capital) and whether an individual is married or cohabiting with their partners. Homosexuals were identified in the sample by cohabitation with a partner of the same sex. Antecol, Jong and Steinberger (2007) found that, after correcting for the amount of hours worked , gay men can expect a penalty in wages (4.5%) when compared to married and a premium (28.2%) when compared to cohabiting heterosexual men. At the same time,
lesbian women can expect a premium in wages when compared to heterosexual women
and the premium is larger in comparison with cohabiting (31.6%) than married heterosexual
women (19.7%). Moreover, they discovered that the premiums were largely explained by
higher educational attainment of homosexual people and to a small extent by differences in occupation. However, the penalty for gay men compared to their married counterparts has remained unexplained.
On top of the studies conducted in the US, several others have been conducted in other
Western countries. Plug and Berkhout (2004) have performed an investigation in the
Netherlands, widely known as a country accepting towards the LGBTQ community, based on a survey of Dutch tertiary education graduates who entered the labour market in years 19982002. The main difference between their study and the earlier ones is that they have identified sexual orientation by directly asking the survey participants about their sexual
orientation, allowing for more precise assessment of nonheterosexuals, inclusion of
individuals who are not in a relationship as well as identification of bisexuals, who previously could have been misidentified as homosexual or heterosexual, and comparing their income
to both heterosexuals and homosexuals. Additionally, rather than studying the sexual
orientation wage gap within gender groups only, Plug and Berkhout (2004) have compared
earnings of bisexual, homosexual and heterosexual men and women simultaneously which
made direct analysis of gender differences possible. On top of that, a distinction has been made between hourly wage and monthly income to correct for disparities in hours worked. According to the results, the monthly wages were about 3% lower for homosexual men and
about 3% higher for homosexual women than for their heterosexual counterparts and, as
Plug and Berkhout (2004) point out, these numbers correspond to the gender wage gap
present in the age group of young people in the Netherlands. Coefficients for bisexuals were not significantly different from zero, which might be because the populations of bisexuals in the sample were too small or because, in the context of earnings, bisexuals are similar to
conducted in France. The studied sample was taken from the Employment Survey, published by the French National Institute of Statistics and Economic Studies and homosexuals were identified through relationship with a person of the same sex. It is also noteworthy that Laurent and Mihoubi (2012) have also investigated for possible disparities in the sexual orientation monthly wage gap between the private and public sector in combination with marriage premiums in both. According to the results, gay men had 6.3% and 5.6% lower
monthly earnings than their heterosexual counterparts in the private and public sectors
respectively while lesbian women monthly earned 2.1% more than heterosexual women in
the private sector and earned the same amount in the public sector. Taking marriage
premium into consideration, the penalties for gay men are 10.2% and 8.1% respectively and
premiums for lesbian women are 3.8% and 0.5%. Furthermore, La Nauze (2015), has
conducted an analysis of the sexual orientation hourly wage gap, based on the Household
Income and Labour Dynamics in Australia survey, where homosexuals were identified
through a relationship with a partner of the same sex. The study also tested whether the sexual orientation wage gap can be explained by personality traits, i.e. whether the widely held beliefs and stereotypes about homosexuals’ nonconformity with the traditional gender roles and characteristics are true and affect their income. The method used to investigate this
is the Big Five classification that measures an individual's extraversion, agreeableness,
conscientiousness, emotional stability and openness to experience. La Nauze (2015) found
that gay men hourly earn between 8% and 18% less than heterosexual men and lesbian
women earn between 0% and 13% more than heterosexual women. Additionally, the
personality traits measures and their effects on earnings are similar across sexual
orientations, meaning that they do not explain the sexual orientation wage gap (La Nauze,
2015). In Greece, Drydakis (2012) compared the hourly earnings of heterosexual,
homosexual and bisexual men and found that gay men earn between 2.9% and 5.9%,
depending on their level of education and whether they are compared with married or
unmarried heterosexuals. The same wage penalty interval for bisexual men is between 3.1%
and 7.5%. In the United Kingdom, Arabsheibani, Marin and Wadsworth (2004) have found
that hourly gay men earn less than heterosexual men, while lesbian women earn more than heterosexual women. Worth noting is the fact that, according to the study, gay men living in London, do not receive a wage penalty and lesbian women’s wage premium is smaller. The same group of researchers has also studied the sexual orientation wage gap with the use of OaxacaBlinder decomposition, in which differences in initial endowments (e.g. abilities) and differences in rewards (e.g. financial rewards based on performance) are treated separately
(Arabsheibani, Marin and Wadsworth, 2005). While, in the study homosexuals of both
genders earned, on average, more than their heterosexual counterparts, the OaxacaBlinder
decomposition has shown that for gay men this premium arises from their, on average,
higher endowments, whereas rewards are lower for them than for heterosexual men. That is,
if homosexual and heterosexual men had the same endowments, the earnings of
homosexual men would be lower. For lesbian women both the endowments and rewards are
higher than for heterosexual women. Lastly, in Sweden, Ahmed and Hammarstedt (2009)
and Ahmed, Andersson and Hammarstedt (2012) found that gay men received an annual
income penalty compared to heterosexual men, while lesbian women received a small and
mostly statistically insignificant wage premium compared to heterosexual women.
III. DATA
The sample studied in this paper has been taken from the American Time Use Survey 2015, published by the US Bureau of Labor Statistics. The dataset has been collected
through Current Population Survey interviews and contains key information about
respondents and members of their households. It has been chosen for the purpose of this study because it includes data on factors necessary for investigating differences in earnings,
particularly sexual orientation wage gap. These factors are: hourly and monthly wages,
annual household incomes and personal characteristics such as gender of household
members and relations between them, age, level of education, ethnicity and region. The sample used has been extracted by excluding observations for which the type of relation was different than ‘reference person living with relatives’, ‘reference person living without relatives’, ‘spouse’, ‘unmarried partner living with relatives’, ‘unmarried partner living without
relatives’ and finally, those whose partner was not present in the remaining dataset.
Homosexuality has been assumed for those respondents who have a spouse or a partner of the same sex in the remaining sample. Moreover, observations with no reported hourly wage have been dropped for the purpose of the analysis of hourly wage wage gap. As a result, 4156 observations, constituting 2078 couples and households, have been obtained. Out of these, only one observation did not have monthly income reported, while all of them had their annual household income reported. For that reason, 4155 observations have been used for the analysis of monthly income and 4156 for the analysis of annual household income. Furthermore, on top of the analyses of three types of income on the comparable samples above, a complementary, additional analysis of annual household income has been conducted due to the fact that the dataset contained many more observations with reported annual household income than hourly wage. The sample used for that analysis has been similarly obtained by dropping observations with a different relation type than the aforementioned ones and those observations for which the annual household income has not been reported. As a result, an additional sample of 25016 observations, constituting 12508 couples has been obtained. Additionally, worth noting is the fact that annual household income was reported as falling into one of several intervals, rather than with particular values. Therefore, for the purpose of this paper it has been assumed to be equal to the middle value of the reported interval, e.g. for the interval of 0 5000, it is assumed that annual household income is 2500. Finally, all earnings are reported in United States dollars.
IV. THEORETICAL FRAMEWORK
In order to test whether there are differences in earnings between homosexuals and
heterosexuals, natural logarithms of hourly wage, monthly wage and annual household
income have been regressed with the use of ordinary least squares method on dummy
variables assigned to gay men, heterosexual women and lesbian women, with heterosexual
men as base. Furthermore, additional variables have been added as controls. Firstly,
variable for age and dummy variables for ethnicity (assigned to Black people, Asian people
and people who identify as neither Black, Asian or White) have been added to the
regression as the basic set of controls (with White people as base). Secondly, variable for the amount of years of education and dummy variables for regions (assigned to Northeast, Midwest and South, with West as base) have been added to the basic set of controls to form
the extensive set of controls . Moreover, standard errors in all the regressions are clustered in response to the fact that all the observations in the sample are coupled in households. Also, as mentioned above, the reason why this paper employs the analysis of three types of income is to study possible differences not only in base wage but also in labour supply and other sources of income across sexual orientations and genders. To be precise, the analysis of monthly income should identify whether such differences do exist as monthly income depends on the amount of hours worked. Furthermore, annual household income depends not only on household members’ base wage and the amount of hours worked but on other sources of income as well. This allows for investigating whether there are additional disparities in this context across studied groups. To summarise, on top of differences in wage, this paper will try to identify differences in labour supply and other sources of income and find if they mitigate or exacerbate the sexual orientation wage gap. Finally, as mentioned above, beside the analyses of the three types of income on a comparable sample, a complementary analysis of annual household income has been conducted to check for the consistency of results.
Therefore, after controlling for aforementioned variables, if there are significant and otherwise unexplained differences in earnings between homosexuals and their heteresexual counterparts, then it could be concluded that there exists a sexual orientation wage gap.
V. RESULTS
Table 1 contains means for all the variables as well as standard deviations. These have been computed by gender and sexual orientation and for the total sample. Means for hourly
and monthly earnings conform with the results of the previous papers on the sexual
orientation wage gap. Namely, on average, gay men earn less than heterosexual men and
women earn less than men in general with lesbian women earning more than their
heterosexual counterparts. However, the differences between average earnings of lesbian
and heterosexual women are modest. Average annual household incomes of heterosexuals
of both genders are very similar, which is to be expected as both heterosexual men and women in the sample are coupled. Annual household income is, on average, lower for gay men than for heterosexuals, while lesbian households earn less than every other group. The average age is similar for all groups and oscillates around 4041 years, with lesbian women
and heterosexual men being, on average, slightly older than the other two groups.
Furthermore, gay men and lesbian women have better education than their heterosexual
counterparts, which is consistent with the existing research. However, the magnitude of this educational advantage is rather small compared to what could be expected. Adding to that,
heterosexual women are better educated than men in general. Ethnicity is similarly
distributed among all groups with two notable exceptions: Black people are
underrepresented among gay men and overrepresented among lesbian women in the
sample. Further, as expected, the biggest fraction of gay men and a significant fraction of lesbian women reside in the West. On the other hand, surprisingly, the biggest fraction of lesbian women resides in the South, while the fractions of homosexuals in the sample who reside in the Northeast are rather small. Finally, in the sample there are 1999 heterosexual
represent about 3% of the male population in the sample, while lesbian women account for 5% of the female population.
Table 2 shows the results of the regression analysis of the natural logarithm of the hourly wages. In column 1, the logarithm is regressed on dummy variables for gay men, lesbian
women and heterosexual women only. Here, the coefficients for lesbian women and
heterosexual women are negative and statistically significant, suggesting wage gap
between these two groups and heterosexual men. The difference in hourly wage for gay men is not statistically significant. In column 2, the basic controls are added. The results are similar to column 1 in terms of statistical significance and the direction of wage gap effect, however, the coefficients are smaller, suggesting that age and ethnicity explain some of the differences found in column 1. In column 3, the extensive set of controls is added, resulting in statistically significant coefficients for all three groups of interest. The regression suggests that gay men face wage penalty of 8.5%, significant at 5% level, compared to heterosexual
men . Furthermore, lesbian and heterosexual women face wage penalties of 15.8% and1
16.9% respectively, compared to heterosexual men, both at 1% significance level. In each of
the columns, lesbian women seem to enjoy a wage advantage, compared to their
heterosexual counterparts, however, in neither case is this advantage statistically significant. Moreover, the fact that all the coefficients in column 3 are statistically significant and larger in magnitude than column 1 suggests that including all the controls allows for a more precise assessment of unexplained wage gaps.
Table 3 shows the results of the regression analysis of the natural logarithm of monthly
wage. In column 1, in which no controls are included, gay men, lesbian women and
heterosexual women seem to receive statistically significant (at 10% level for gay men and 1% level for both groups of women) wage penalties compared to heterosexual men. When
adding the basic controls, significance levels stay the same, while the magnitudes of
differences shrink. When controlling with the extensive set of variables, the wage penalties become again larger. Namely, gay men can expect to receive 18.9% lower monthly wages, compared to heterosexual men. This effect is statistically significant at 5% level.
Furthermore, lesbian and heterosexual women receive a monthly wage penalty of 28.5%
and 31%respectively, when compared to heterosexual men, both statistically significant at
1% level. Finally, homosexual women seem to earn more than heterosexual women in all three regressions, however, this difference is very small and not statistically significant in either case. The wage penalties for all three groups of interest are about twice as large for monthly income as the penalties in hourly wage.
Table 4 contains the results of the regression analysis of the natural logarithm of the annual household income. In column one no controls are present and both gay men and lesbian women receive an earnings penalty that is statistically significant at 5% and 1% level respectively, when compared to heterosexual men. The difference in household earnings for
heterosexual women is insignificant. When adding the basic controls, the penalties for
homosexuals become slightly smaller and the coefficient for gay men is significant only at
10% level. Furthermore, the difference in earnings for heterosexual women is slightly
positive and becomes significant at 1% level. Including the extensive set of controls in the regression results in statistically significant differences in earnings for all three groups of interest. Namely, gay men can expect 24% lower household income than their heterosexual counterparts. This is significant at 5% level. Moreover, lesbian women can expect 27% lower
household income, compared to heterosexual men, significant at 1% level. Additionally,
heterosexual women can expect 2.1% lower household income than heterosexual men,
significant at 5% level. Finally, when including all the controls, lesbians obtain 25.2% smaller household income than heterosexual women and this is significant at 1% level.
Table 5 contains means and standard deviations of all the variables used for the complementary analysis of annual household income. Annual household income exhibits the same tendencies as in the sample used above, with heterosexual households earning the most and lesbian households earning the least. However, with the exception of lesbian households, all households in this sample are, on average, more affluent. Furthermore, individuals in this sample are somewhat older than in the previous one, with the average age oscillating around 47. In terms of ethnicity, the results are similar to the sample above. Worth noting is the fact that the Black population is overrepresented among lesbian women and is not underrepresented among gay men as was the case in the sample above. As to education, gay men are the most educated groups in this sample, which is consistent with past research. However, lesbian women are the least educated group, which is rather surprising. Moreover, the biggest fraction of each groups resides in the South, which is not expected for homosexuals. Additionally, considerable fractions of homosexuals reside in the West, which conforms with the expectations. Finally, in the sample there are 114 gay men and 342 lesbian women. As there are 12280 heterosexual men and 12280 heterosexual women, gay men represent about 1% of the male population, while lesbian women represent about 3% of the female population. Therefore, compared to the previous sample, populations of homosexuals are rather small and there are 3 times as many lesbian women as there are gay men.
Table 6 contains the results of the complementary regression analysis of the natural logarithm of the annual household income on a bigger sample. In column 1, with no controls, the result is statistically significant only for lesbian women and it is quite large. When adding the basic controls in column 2 it becomes smaller, but the results for gay men and heterosexual women remain statistically insignificant, with the coefficient for heterosexual women being very close to 0. In column 3, the results for lesbian women and heterosexual women are statistically significant at 1% level. They can expect 31.8% and 1.2% lower household income respectively, when compared to heterosexual men. Furthermore, gay men can expect 11.8% lower household income than heterosexual men, however, this result is not statistically significant. Finally, lesbian households’ income seems to be 30.9% lower than for heterosexual women and this is significant at 1% level.
VI. DISCUSSION
The outcomes of hourly wage and monthly regressions are consistent with the existing
research. Namely, gay men can expect significantly lower earnings, compared to
heterosexual men. At the same time, women in general earn less than men and lesbian women earn slightly more than heterosexual women, albeit this premium is not statistically significant. Therefore, there exists an unexplained sexual orientation wage gap for gay men, while the difference between earnings of lesbian women and heterosexual men seems to be fully explained by the gender wage gap. Worth noting is the fact that the magnitude of differences in earnings between gay men, women in general and heterosexual men is more or less doubled, when comparing monthly earnings with hourly wage. Although the amount of hours worked was not included in the analysis in this paper, it is likely that the monthly wage differentials are higher than those of hourly wage because of differences in hours worked per month across the studied groups. This would mean that in general gay men, lesbian women and heterosexual women are all more likely to work parttime. This could
particularly distort the results if homosexuals and heterosexual women were more likely to work parttime than heterosexual men.
The results of the first annual household income regression analysis show different trends than the previous two. Namely, heterosexual women earn only slightly less than heterosexual men. This, however, can be explained by the fact that heterosexual women in the sample are paired with heterosexual men and form households together. One could argue that the outcomes for gay men and lesbian women in this analysis are consistent with the previous studies as the former earn less than heterosexual men and the latter earn less than men in general, because in both cases each household member faces wage penalties (sexual orientation wage gap and gender gap respectively). However, in this particular regression, the assumption of household income being equal to the middle value of the given household income interval can result in major distortions and thus unreliability. Therefore, one should be careful when trying to draw any conclusions here.
Moreover, the complementary analysis of the annual household income with the use of a bigger sample shows similar, albeit slightly higher, penalty for lesbian women, compared to both heterosexual men and heterosexual men, thereby proving the consistency of the trend for lesbian women households. However, the penalty for gay men is less than half of the one in the initial annual household income analysis and the result is not statistically significant. This might be due to the possible small sample bias as in this analysis gay men represent a very small proportion of the male population in particular and the whole sample in general.
Furthermore, the observed unexplained sexual orientation wage gap for gay men does not imply that there is discrimination based on sexual orientation. Firstly, this might be due to occupational sorting, that is, if gay men choose to work in professions that are considered more feminine or more gayfriendly and are often less paid, they will observe lower earnings, similarly to heterosexual women (Antecol, Jong and Steinberger, 2007). It is further proven by Plug, Webbink and Martin (2014) that both gay men and lesbian women prefer to work in
professions that are more friendly towards homosexuals, suggesting that occupational
sorting might have a major effect on the existence and magnitude of the sexual orientation
wage gap. Secondly, the possible existence of marriage premium could also explain the
wage gap. The marriage premium is the unexplained positive effect of being married on heterosexual men’s earnings and it has been proven to further widen wage gaps in studies of earnings of sexual orientation minorities (Laurent and Mihoubi, 2012). Thirdly, although La Nauze (2015) has not found any evidence to support the theory that personal characteristics explain the sexual orientation wage gap, there might be differences between Australia and the US in terms of attitudes towards homosexuals’ nonconformity with the traditional gender roles, that could allow for these characteristics to partially explain the wage gap.
On top of that, as mentioned above, the increased magnitude of the penalties in monthly earnings for gay men, lesbian women and heterosexual women, compared to hourly earnings, shows strong likelihood of differences in labour supply. Furthermore, lesbian women’s hourly and monthly wages are not significantly different from heterosexual women’s wages, but all women’s hourly and monthly wages are significantly lower than for heterosexual men. Therefore, it seems that gay men exhibit characteristics somewhat
different from heterosexual men, while lesbian and heterosexual women exhibit similar characteristics. This could be explained by the fact that homosexual men are less competitive than heterosexual men, while lesbian women do not show signs of competitiveness different from heterosexual women (Buser, Geijtenbeek and Plug, 2015). This result would be consistent with the fact that there are differences in earnings and labour supply between homosexual and heterosexual men and with the fact that there are no such differences among women (i.e. lesbian women do not receive premiums or penalties on top of those that can be expected by women in general). The above arguments do not contradict the results of the annual household income analyses as the household incomes of heterosexual women include earnings of their heterosexual male partners whose higher earnings compensate for the gender gap.
Moreover, there are several caveats that could potentially limit the internal and external validity of this analysis. Firstly, the populations of gay men and lesbian women in the sample are relatively small and could result in a small sample bias. Secondly, it has been proven in
past research that homosexuals are, on average, much more educated than both
heterosexual men and women. This is also the case in this paper (with the exception of the
average education among lesbian women in the second sample), however, the educational
advantages of homosexuals are not as high as it would be suggested by previous research. Therefore, the populations of homosexuals might not be fully representative of the sexual minority population in the US. Thirdly, the sample used in this paper consists only of individuals who are in a relationship or are married and, therefore, it is impossible to assess
the differences between unmarried, married, single and cohabiting individuals, such as
marriage premium. Lastly, it is possible that identification of homosexuality through
relationship with a samesex partner is not a fully reliable method and could result in distortions in the outcomes
Lastly, based on the aforementioned points, several suggestions are made for further
research. To begin with, new data collection methods should be developed, allowing for a more precise assessment of sexual orientation. This is becoming more possible with the
development of technology, while changing attitudes towards the LGBTQ community make
social selectivity bias less likely. Further, a comprehensive analysis of all the factors that potentially could explain the sexual orientation wage gap should be conducted. Proposed additional controls for this analysis are: occupation, relationship status, marital status, personality characteristics (e.g. Big Five), information about whether sexual orientation has been disclosed to the employer. Additionally, comparing the size of the sexual orientation wage gap across time could provide valuable insights for the academic discussion of the problem, especially in the context of the US, which as of 2015 has introduced samesex marriage countrywide. Particularly, this would also allow for testing whether married gay men also receive a marriage premium. Finally, comparing the results of sexual orientation wage gap studies from different countries could also provide new information on the topic such as the effects of different policy measures or social attitudes on sexual orientation minorities’ earnings.
VII. CONCLUSION
In this study, the existence of the sexual orientation wage gap has been investigated. It has
been found that gay men can expect 8.5% lower hourly wage and 18.9% lower monthly
wage than heterosexual men, ceteris paribus, and these differences in earnings are
statistically significant. Furthermore, lesbian women do not receive significantly different earnings than heterosexual women, however, all women can expect their hourly wage to be about 16% lower and their monthly wage to be about 30% lower than heterosexual men, ceteris paribus. Therefore, there exists an unexplained sexual orientation wage gap for gay men but there is no such evidence for lesbian women. Instead, the differences in earnings of
lesbian women can be mostly explained by the gender wage gap. Thus, there is no
evidence that all homosexuals are discriminated in wages. Finally, the results of this study do
not warrant inferences about whether discriminations is behind the unexplained sexual
orientation wage gap for gay men. Further studies would need to be conducted to test the validity of such claims.
REFERENCES
Ahmed, A., Hammarstedt, M. (2010). S exual orienta on and earnings: a register data‐based approach to iden fy homosexuals. Journal of Population Economics, 23 (3), 835‐849.
Ahmed, A., Andersson, L., Hammarstedt, M. (2013). Sexual orienta on and full‐ me monthly earnings, by public and private sector: evidence from Swedish register data. Review of Economics of the Household, 11 (1), 83‐108. Allegre o, S. A., Arthur, M.M.(2001) An Empirical Analysis of Homosexual/ Heterosexual Male Earnings Differen als: Unmarried and unequal? Industrial and Labor Relations Review , 54 (3), 631‐646. Antecol, H., Jong, A., Steinberger, M. (2008). The Sexual Orienta on Wage Gap: The Role of Occupa onal Sor ng, Human Capital, and Discrimina on. Industrial and Labor Relations
Review, 61 (4), 518‐543.
Arabsheibani, G. R., Marin, A., Wadsworth, J. (2004). In the pink: Homosexual‐heterosexual wage differen als in the UK. International Journal of Manpower, 25 (3/4), 343‐354.
Arabsheibani, G. R., Marin, A., Wadsworth, J. (2005). Gay Pay in the UK. Economica, 72 (286), 333‐347. Badge , M. V. L. (1995). The Wage Effects of Sexual Orienta on Discrimina on. Industrial and Labor
Relations Review, 48 (4), 726‐739.
Black, D., Gates, G., Sanders, S., Taylor, L. (2000). Demographics of the gay and lesbian popula on in the United States: Evidence from available systema c data sources. Demography, 37 (2), 139‐154.
Blandford, J. M. (2003). The Nexus of Sexual Orienta on and Gender in the Determina on of Earnings. Industrial and Labor Relations Review , 56 (4), 622‐42.
Buser, T., Geijtenbeek, L., Plug, E. (2015). Do gays shy away from compe on? Do lesbian compete too much? Institute for the Study of Labor (IZA)
Coffman, K. B., Coffman, L. C., Marzilli Ericson, K. M. (2016). The Size of the LGBT Popula on and the Magnitude of An gay Sen ment Are Substan ally Underes mated. MANAGEMENT SCIENCE
Articles in Advance , 1–19.
Denney, J. T., Gorman, B. K., Barrera, C. B. (2013). Families, Resources, and Adult Health.
Journal of Health and Social Behavior 54 (1), 46‐63.
Drydakis, N. ( 2012). Sexual orienta on and labour rela ons: new evidence from Athens, Greece.
Applied Economics 44 (20), 2653‐2665.
La Nauze, A. (2015). Sexual orienta on–based wage gaps in Australia: The poten al role of discrimina on and personality. The Economic and Labour Relations Review , 26 (1),60‐81. Laurent, T., Mihoubi, F. (2012). Sexual Orienta on and Wage Discrimina on in France: The Hidden
Side of the Rainbow. Journal of Labor Research, 33 (4), 487‐527. Liu, H., Reczek, C., Brown, D. (2013). Same‐sex cohabitors and health: the role of race‐ethnicity, gender, and socioeconomic status. Journal of health and social behavior 54 (1), 25‐45. Plug, E., Berkhout, P. (2004). Effects of sexual preferences on earnings in the Netherlands. Journal of Population Economics, 17 (1), 117‐131. Plug, E., Berkhout, P. (2008). Sexual Orienta on, Disclosure and Earnings. Institute for the Study of Labor (IZA)
Plug, E., Webbink, D., Mar n, N., (2014). Sexual Orienta on, Prejudice, and Segrega on. Journal of
Labour Economics, 31 (1), 123‐159.
APPENDIX FULL REGRESSION TABLES