Over the past decade, the Census Bureau has made changes to address data quality issues and allow for the identification of same-gender relationship households and in the September 2020 release of the American Community Survey, the Census Bureau disaggregated data for these families. This release marks the first-time researchers could look at federally-collected nationally-representative estimates of the number of same-gender relationship households in a post-Windsor world, where same gender marriage legalized in all states. These changes are instrumental as a first step in helping researchers and policymakers understand the circumstances of same-gender  relationships from these data. (See the technical appendix for detailed discussion on previous estimates).
Because these data are so new, we make two contributions that are intended to support our own and others’ future work. First, building off research released by Census, we examine the economic status of same-gender households: to be clear, we are only able to identify same-gender relationship households who are married or partnered and co-resident. This analysis takes a deeper dive into the median incomes of same-gender families and looks at their compositional differences to better understand why some same-gender families have higher median incomes than opposite-gender families. Second, to speed future research we develop a detailed technical appendix that explains the history of Census’ misclassification of same-gender relationship households and how to use these data to study these families.
Family Income and Family Structure
Since making information on same-gender households available, the Census Bureau has released multiple reports providing demographic data on same-gender married couples. They find that, on average, same-gender married couples have higher median household incomes and higher rates of dual employment than opposite-gender married couples. We extend the Census Bureau’s analysis with the following differences. We separate same-gender families into those with two male and two female partners, as we would expect their labor market experiences and incomes to differ from each other and that of different-gender couples. We limit income to the family rather than household, that is, only the reported income of the respondent and their spouse or partner are included. We found that same-gender households were more likely to have additional earners in the household outside of the focal partnership than opposite-gender households; this affected household income. We also move from the single-year 2019 ACS to the five-year 2015-2019 ACS.
In the Census Bureau’s report, they found that, on average, the median household income of same-gender households is $107,200 compared to $97,000 for opposite-gender married couples. We replicate this finding in the five-year file. When disaggregated by the couple’s genders and marital status, we find that the median incomes of male and female same-gender couples are quite different from each other. Adult men in same-gender couples have the highest family incomes regardless of marital status. On average, the family income for married men in same-gender relationships is 31 percent higher than married women in same-gender relationships, and 27 percent higher than opposite gender married couples. The income gap for men in unmarried partnerships is 36 percent higher than unmarried women in same-gender relationships, and 38 percent higher than opposite gender unmarried couples.
Perhaps surprisingly, women in a same-gender coupled family – regardless of marital status—have similar family income to opposite-gender couples. On average, same-gender female unmarried coupled families have an average income that is about $2,000 higher than opposite-gender unmarried coupled families and $3,000 lower than opposite-gender married couple families.
Income-Relevant Differences Between Families
Prior research suggest that presence of children, education, being in a dual income household, and residing in a high-density area are associated with higher household income. Compositional differences by household structure along these dimensions could help explain differences in family income. Figure 2 shows the share of prime-age opposite-gender couples (green bar), same-gender female couples (yellow box), and same-gender male (orange circle) that report a child in the household, the presence of a bachelor’s degree holder, having two earners, and living in the highest quintile of population density.
Married couples are more likely to have a child present, are more highly educated, are more likely to live in a high-density place, and are less likely to have two earners. Same-gender couples are less likely than opposite-gender couples to have a child present, but are more likely to have two earners, be more highly educated, and live in a densely populated place. Notably, same-gender male couples are much more likely to live in a high density area, have a bachelor’s degree holder in the relationship, and less likely to have a child than other households. While these results are perhaps intuitive, it is nevertheless the first time that federally-collected nationally-representative data have allowed for this analysis.
To further investigate the relationship between these factors and family income among prime-age families, we estimate an ordinary least squares (OLS) regression model. We regress family income on age, education, couple type, the number of earners, population density, and the presence of children without addressing the gender makeup of the couple (model 1), then add whether the couple is of the same gender or not (model 2) and then finally delineate the following couple types in model 3:
- same-gender male married
- same-gender female married
- opposite-gender unmarried cohabitants
- same-gender male unmarried cohabitants
- same-gender female unmarried cohabitants
- male singletons
- female singletons.
The baseline relationship for this model (the excluded group) is an opposite-gender married couple. Results from these regressions can be found in the technical appendix (Appendix Table 2).
This analysis finds that some, but not all, of the income gaps described above are associated with other characteristics associated with income. Before accounting for other characteristics, same-gendered male married couples earn $18,000 more than different gendered married couples and male unmarried couples earn $33,000 more than different gendered unmarried couples. Holding constant the characteristics described above, those positive income gaps are $27,000 and $11,000, respectively. On the other hand, before accounting for other characteristics, same-gendered female married couples earn $3,000 less than different gendered married couples and female unmarried couples earn $6,000 more than different gendered unmarried couples. Holding constant the characteristics described above, both income gaps are negative: $11,000 and $26,000, respectively.
To anchor our understanding of how the characteristics are associated with income, assume there is a relationship comprised of two prime-aged adults of any gender (model 1). Families where one adult earned a bachelor’s degree or higher are associated with having incomes that are on average $67,000 higher than families where neither adult has such a degree, all else constant. Families with two earners in the home are associated with having incomes that are on average $33,000 higher than relationships that have one earner (or no earners), all else constant. We find that families with children are associated with having incomes that are on average $2,500 higher than families without children, all else constant. Families who live in the highly density areas are associated with having incomes that are on average $11,000 higher than those who don’t live in highly dense areas, all else constant. We find that families with two married adults are associated with having incomes that are $33,000 higher than families with one single adult, all else constant. Families where the adults are unmarried partners are associated with having incomes that are $12,000 higher than families with a single adult present, all else constant. The first model shows us that there is a clear association between the compositional factors of families, marital status, and their family income.
In model 2, we add an indicator for same-gender married and partnered couples regardless of gender to the model to see if the pattern holds. Holding everything else constant, we find that families with same-gender couples are associated with having incomes that are $8,500 higher than families with opposite gender couples, all else constant.
Finally, we fully specify couple types by both marital status and the gender composition of the couple, relative to opposite-sex married couples. All else constant, families with a same-gender male couple, both married and partnered, are the only couple type associated with having higher family incomes than opposite-gender married couples: $27,000 and $11,000 higher, respectively. Unlike their male counterparts, same-gender married female coupled families are associated with having average incomes that are $11,000 lower than opposite gender married coupled relationships, all else constant and women in same-gender unmarried coupled families are associated with having average incomes that are even lower: $26,000. Families with an opposite gender unmarried couples are associated with having incomes that are on average $21,000 less than opposite gender married couples, all else constant.
Regressing couple type on family income reveals stark differences in the relationship between marital status, the genders of the members of the couple, and income. Men in same-gender couples have the highest proportions of combined characteristics associated with higher family incomes and the highest earners. Women in same-gender couples are more likely to have two earners, higher education, and live in a densely populated area than opposite-gender couples, but their family incomes are the lowest of the couple types.
While we offer a first look at the family incomes and compositions of same-gender couples, there are limitations to our study. The American Community Survey does not ask questions about sexual orientation; we can only identify the share of adults who are coupled and cohabitating with a spouse or partner of the same gender. We are not able to address other factors that impact income which we do not observe, including discrimination by gender, gender identity, or sexual orientation; those who identity as LGBTQ+ report experiences of workplace discrimination which include being fired, passed up for employment opportunities, or experiencing harassment resulting from their sexual orientation. To be sure, many of the explanatory factors regarding income are not accounted for in these models, most notably industry, occupation, and tenure. These factors affect the economic status of LGBTQ-identified adults and same-gender relationship families.
There are many factors that contribute to differences in family income. For the first time, nationally-representative data of a post-Windsor world allow researchers to look at family income of same-gender couples who are married or partnered. We find that overall, while men and women in same-gender families have similar proportions of income-related characteristics and higher proportions than opposite-sex couples, men in same-sex relationships have by far the highest median incomes.
This technical appendix accompanies Examining the Economic Status of Same-Gender Households. This analysis describes the demographics of households, including same-gender relationship households, using national federally-collected survey data: the American Community Survey (ACS) 2015-2019 five-year file. We find that adults in same-gender relationships, particularly men, have different demographic profiles from adults in opposite-gender relationships. The technical appendix also serves as a primer for those looking to use the ACS identification of same-gender households.
The Brookings Institution is financed through the support of a diverse array of foundations, corporations, governments, individuals, as well as an endowment. A list of donors can be found in our annual reports published online here. The findings, interpretations, and conclusions in this report are solely those of its author(s) and are not influenced by any donation.
The authors thank Wendy Edelberg and Este Griffith for helpful feedback. Sara Estep, Nidhi Nair, Elisabeth Raczek, and Natalie Tomeh provided excellent research assistance. The authors are particularly grateful to Gary Gates for providing feedback on the technical appendix.