Wealth, gender and sexual orientation— evidence from siblings

Using Swedish administrative data, this study investigates the link between wealth and sexual orientation across genders, focusing on nearly 4400 individuals who have ever been in a same-sex legal union and their siblings who had been exclusively in different-sex relationships. Employing unconditional quantile regressions and sibling fixed effects, we show that the wealth gap by gender and sexual orientation varies across the wealth distribution. Men in same-sex couples (SSCs) experience a wealth penalty below the 70th percentile but a premium above it. For women, the wealth penalty persists until the 95th percentile. Similar patterns hold for the wealth subcomponents, with men in SSCs holding more financial resources, real estate and debt at the top of the distributions, while women in SSCs hold more financial resources but less real estate and total debt. Additional analysis highlights the positive marginal effects of urban residency and years of schooling on these patterns.


Introduction
Women typically have lower labor income and, consequently, less wealth than men (e.g.Warren, 2006;Sierminska et al., 2010;Ruel and Hauser, 2013;Schneebaum et al., 2018).While a substantial body of literature documents a significant wealth gap for racial-ethnic minorities and women compared with white men (Neelakantan and Chang, 2010;Schneebaum et al., 2018;Pugliese and Belleau, 2021;Bessi� ere, 2022;Pugliese and Belleau, 2022), less is known about how wealth varies by sexual orientation.A major reason for this research gap is the scarcity of administrative wealth data that include information on sexual orientation, which makes it difficult to examine the wealth disparities experienced by sexual minority individuals.Addressing this research gap is important given that wealth plays a crucial role in people's well-being in terms of financial security, health, augmented revenues and national Gross Domestic Product (GDP) growth (Wolff, 1998).Additionally, the considerable and increasing size of the sexual minority population (OECD, 2019;Badgett et al., 2021) underscores the need for a deeper understanding of wealth accessibility across gender lines among sexual minorities.
In this article, we study the wealth gap by sexual orientation and gender using Swedish administrative data for the entire population.Sexual orientation is defined as whether an individual has ever been in a same-sex legal union or has been exclusively in different-sex legal unions and our definition of 'gender' is a binary variable based on legal gender.Sweden is a country with a history of openness toward sexual minorities.According to data from 2017 to 2020, in Flores' (2021) ranking of 175 countries, Sweden is the fourth most accepting country for LGBTQ people behind Iceland, the Netherlands and Norway.Consistent with this, Sweden legally recognized same-sex unions and established a registered partnership in 1995.This move made Sweden the third country in the world to do so, following Denmark and Norway.Registered partnerships in Sweden are granted most of the same rights and responsibilities as full legal marriages.After the introduction of registered partnerships in 1995, gender-neutral marriage legislation was introduced in 2009 (Ald� en et al., 2015).Consistent with Kolk and Andersson (2020), this study employs the term 'legal unions' to refer to both registered partnerships and marriages.
Using sibling fixed effects (FEs), we address potential selection biases in our identification of sexual orientation, as well as confounding factors such as family background and values, childhood geographical characteristics, parental educational attainment and labor market outcomes (e.g.Brattbakk and Wessel, 2013;Humlum et al., 2019;Ald� en and Neuman, 2022).In fact, Rothblum et al. (2004) find that heterosexual individuals with sexual minority siblings are comparable to heterosexual individuals in the general population of the same age, including in terms of marital status, childbearing and religious values.By restricting our sample to families with at least two siblings of the same gender but with different sexual orientations, we estimate the difference in wealth gap depending on gender and sexual orientation.
Using unconditional quantile regressions (UQRs), we show that the wealth gap by gender and sexual orientation varies significantly over the wealth distribution.Although men who have ever been in a same-sex couple (SSC) possess more net wealth than their brothers who have been exclusively in different-sex couples (DSCs) at the 90th percentile (P90), men who have ever been in an SSC have less wealth than their DSC brothers at the median (P50).Sexual minority women tend to have less wealth than heterosexual sister(s) over the entire net wealth distribution, but these negative wealth gaps vary in magnitude over the net wealth distribution.Additionally, when analyzing the components of wealth, we observe interesting patterns in the distribution.Finally, we consider two potential mechanisms to explain wealth differences among siblings by gender and sexual orientation: education and rural/urban channels.
Our research makes several contributions to the literature.First, to our knowledge, this article is the first to examine the relationship between net wealth and sexual orientation by gender.Second, we benefitted from using much larger and higher-quality data sets than did any previous study.A larger sample size allows greater precision in estimating the independent relationship between sexual orientation and wealth.Our wealth registry data are administrative in nature and, thus, have higher reliability than self-reported wealth from surveys and cover the entire Swedish population (Roine and Waldenstr€ om, 2009;Betermier et al., 2017;Coffman et al., 2017;Bach et al., 2020;Boschini et al., 2020).Third, by using a sibling approach in our analysis, we account for shared family background, values and upbringing, which allows us to focus on the effect of sexual orientation on wealth outcomes (e.g.Perales and Huang, 2020;Reczek et al., 2023).Fourth, our study explores several economically interesting outcomes not addressed in previous work, such as geographical interplay, liabilities and financial and real estate asset composition across sexual orientation by gender.Overall, our contributions to research on wealth and sexual orientation by gender are aligned with the growing literature on the wealth of marginalized groups based on race (Wolff, 1998;Charles and Hurst, 2002;McKernan et al., 2014), gender (Neelakantan and Chang, 2010;Ruel and Hauser, 2013;Schneebaum et al., 2018;Merik€ ull et al., 2021) and the intersection of race and gender (Warren, 2006;Brown, 2012).

Literature review
Following Badgett's seminal work in 1995 on wage differences between heterosexual and sexual minority men and women in the USA, there has been substantial growth in the literature on the economics of sexual orientation (Badgett, 1995).Based on previous studies, there are several reasons to expect wealth to differ across sexual orientations by gender.A significant body of literature has identified gender earnings disparities between sexual minority individuals and their heterosexual counterparts.Empirical studies in this field have revealed a recurring pattern in which lesbian women earn more than comparable heterosexual women do, while gay men earn less than comparable heterosexual men even after controlling for variables such as years of schooling, age, experience, marital status, childbearing and other relevant factors (Ahmed et al., 2011;Ald� en et al., 2015;Klawitter, 2015;Badgett et al., 2021).While there have been instances in the academic literature where terminology such as 'gay', 'lesbian' and 'bisexual' have been used to distinguish between sexual orientations, in our article, we utilize the term 'sexual minority individuals' to refer specifically to those who have ever been in SSCs.
Homeownership is often people's largest source of wealth (e.g.Fligstein and Goldstein, 2015;Kuhn et al, 2020;Kuypers et al., 2021;Merik€ ull et al., 2021).Previous research has investigated differences in homeownership rates and self-reported values across sexual orientations (Leppel, 2007;Jepsen and Jepsen, 2009;Conron et al., 2018;Pugliese and Belleau, 2022).Understanding these disparities is crucial for identifying structural advantages or disadvantages based on sexual orientation by gender, given the personal and social benefits associated with homeownership.Previous studies have shown that SSCs may resort to homeownership to bypass leasing discrimination, the likelihood of which varies depending on the national context.Compared with DSCs, males in same-sex relationships have a 20-30% lower likelihood of being approved in the American rental housing market, whereas females in same-sex relationships, on average, experience a 10% lower likelihood (Ahuja and Lyons, 2019;Gouveia et al., 2020).
However, several factors may hinder homeownership and capital access.
LGBTQ marriage legislation constraints (Miller and Park, 2018;Sansone, 2019;Dilmaghani and Dean, 2020) and fragile labor market outcomes (Ozeren, 2014;Klawitter, 2015;Drydakis, 2022) may have detrimental effects on homeownership.Using the US couple data census, Leppel (2007) and Jepsen and Jepsen (2009) reported that SSCs are less likely to own a home than heterosexual couples are, even after controlling for observable characteristics such as income, educational attainment and fertility.
Access to capital and credit is a key driver of homeownership and wealth; therefore, it can be an important factor in explaining wealth disparities.Discrimination in the capital market can impede access to credit and capital, particularly for marginalized groups (Leppel, 2007;Jepsen and Jepsen, 2009;Miller and Park, 2018;Dillbary and Edwards, 2019).Negrusa and Oreffice (2011) and Sun and Gao (2019) provided evidence of credit market discrimination against sexual minorities in the USA, with SSCs facing higher mortgage refusal and interest rates than their heterosexual peers, despite exhibiting similar financial and credit quality.Using more than five million mortgage applications, Dillbary and Edwards (2019) found that same-sex male co-applicants are more likely to be denied than are same-sex female applicants and different-sex applicants.
Risk preferences might be another factor impacting access to wealth.While early literature, for example, Eckel and Grossman (2008) and Croson and Gneezy (2009), found women to be more risk averse than men, more recent insights suggest that exactly how risk is elicited and measured might impact potential gender gaps in risky behavior.See, for example, Boschini et al. (2019), Alam et al. (2022) and Comeig et al. (2022).However, Buser et al. (2018) find no risk aversion differences between sexual minority individuals and their heterosexual counterparts.
The geographical location of residence may serve as another explanatory factor for wealth inequalities based on sexual orientation.Urban areas are characterized by greater human capital, income, wealth and economic opportunities (Black et al., 2002;Florida et al., 2008;Florida, 2019).Studies have shown that gay men are more likely than lesbian women and heterosexual individuals to live in urban areas with more tolerant and progressive environments (Black et al., 2000;Gates and Ost, 2004;Rothblum et al., 2004;Klawitter, 2011;Denier and Waite, 2018).According to Badgett et al. (2021), individuals in SSCs are more likely than those in DSCs to reside in a US state other than their state of birth and to have moved to regions with higher salaries, greater urbanization and more progressive attitudes.
Finally, while there are only a few studies on the differences in educational attainment and financial literacy across sexual orientations, their findings suggest that sexual minorities tend to have higher levels of education and prefer to major in fields characterized by less prejudice (Andersson et al., 2006;Ahmed et al., 2011;Sansone and Carpenter, 2020;Badgett et al., 2021) and more workplace independence (Burn and Martell, 2020).Thus, it is plausible that variations in financial literacy, specialization and educational attainment could contribute to wealth disparities across sexual orientations (Altmejd et al., 2022).

Data
To examine wealth outcomes, we use the Swedish wealth register, which contains individual-level wealth data for the entire Swedish population.The Swedish wealth register existed from 1999 and was discontinued after the wealth tax repeal in 2007, as financial institutions were no longer obliged to report individuals' assets and liabilities to the tax authorities.The wealth register contains third-party reported variables on various real and financial assets and debt at the personal level.All the variables are in real values (with base year 1999) and are denominated in SEK.The individual high-quality data reported by financial institutions, banks and other public entities to the national tax agency contrasts with the often self-reported survey data and household wealth data used elsewhere.For more details on the Swedish wealth register, see, for example, Roine and Waldenstr€ om (2009), H€ allsten and Pfeffer (2017) and Lundberg and Waldenstr€ om (2018).Our main wealth variable is yearly individual net wealth, which is defined as the sum of the total market value of financial assets (i.e. bank account balance; interest, mixed and equity funds; shares, quoted options; bonds; domestic and foreign endowment insurance; and securities) plus the sum of real estate assets (i.e.houses, agricultural, industrial, domestic, foreign and land properties) minus the total market value of the individual's debt (i.e.mortgage and student debt).To understand the source of differences in net wealth between individuals, we also use the components of net wealth-financial assets, real estate assets, student debt and total debt-separately.
In addition, we use the Longitudinal Integrated Database for Health Insurance and Labor Market Studies (LISA), which contains administrative information covering a range of demographic and socioeconomic factors at the yearly individual level for the entire Swedish population over 16 years from 1990 onward.More specifically, from the LISA, we use information about the individual's age, civil status, years of schooling, earnings, whether the individual lives in an urban or rural area, whether the individual has any siblings, and the legal gender of the individual(s) and of their sibling(s).Sweden enforces a binary legal gender system, where legal gender is primarily determined on the basis of the sex assigned at birth.Sweden does not recognize third-gender or non-binary identities.This legal framework creates a potential discrepancy between legal gender and gender identity, which limits the identification of gender minority individuals.For the purposes of this article, 'gender' refers to the legal gender as recorded in the registry data.
Using the LISA, we are able to determine whether individuals older than 18 years who legally reside in Sweden have ever been in an SSC (either a registered partnership or marriage) or been exclusively in DSCs.For most of our sample period, LISA lacks information on cohabitating couples unless they share custody of their children.This limitation arises due to a significant asymmetry in the prevalence of children among same-sex male and female couples, which is particularly noteworthy because, until 2018, only SSCs in a legal union were eligible to adopt.Although data on shared living arrangements became available in 2011, the ambiguity of such shared housing units-including housemates, close relatives or distant family members-precludes their use in reliably determining an individual's sexual orientation.Given that all our data stemmed from administrative population registers, we lack other information on individuals' sexual orientation.

Sample restrictions
In 1995, a registered partnership for SSCs was introduced in Sweden.This legal union status grants similar rights to marriage (except for the opportunity to adopt a child until 2003), access to medically assisted insemination (until 2005), and the requirement of being a legal resident before entering into a registered partnership (Rydstr€ om, 2011;Kolk and Andersson, 2020).Gender-neutral marriage legislation was enacted in Sweden in 2009, leading to the cessation of new registered partnerships.Couples existing in a registered partnership before the legalization of same-sex marriage were provided the choice to either convert their registered partnership into a marriage or retain their registered partnership status, as opposed to marriage, according to their preference.In this study, we follow Kolk and Andersson (2020) and use the term 'legal unions' to refer to both registered partnerships and marriages.To ensure the accuracy of our identification of sexual orientation, we excluded individuals who had never been in a legal union from 1995 to 2020 following the same methodology devised in the literature (Andersson et al., 2006;Ahmed et al., 2011;Ald� en et al., 2015;Kolk and Andersson, 2020).
Since 1995, the share of same-sex unions has increased annually, while the number of different-sex unions has remained roughly stable; see Figure 1-see Appendix Figure A1, for the age distribution of the first legal union across the full Swedish population .
Utilizing comprehensive data spanning from 1995 to 2020 to determine individuals' sexual orientation, while possessing wealth data solely for the period 1999-2007, introduces inherent age asymmetry.Consequently, individuals in SSCs within our sample are likely to be younger than their counterparts in DSCs during the years covered by the wealth data.Our study included individuals who maintained a legal relationship for a minimum of 1 year and excluded 3 million individuals who had never participated in any form of legal union.The reference group for comparison comprises individuals who were exclusively observed to be in a different-sex legal union.Approximately 8% of the individuals who were ever in an SSC had also been in a DSC at some point in their life.Thus, we include them in the 'ever in an SSC' variable for a comprehensive analysis.
Panel A in Figure 2 shows the age distribution by gender and sexual orientation for the full population, and it is especially the case that women who have ever been in an SSC tend to be considerably younger than women exclusively in a DSC in the years for which there are wealth data.Since wealth accumulates over individuals' lifetimes, the age skew in the sample risk is overestimating the wealth gap by sexual orientation and gender.To address this skewness, we restrict our analysis to include only siblings of the same gender and biological parents in families where there is at least one sibling who enters a same-sex union and at least one sibling who enters a different-sex union.Panel B in Figure 2 indicates that this approach notably reduces skewness with respect to age.Table 1 presents the share of individuals belonging to different wealth quantiles by gender and union type, as well as the average age and years of schooling in 2003.This is presented separately for the full population and for the sibling samples.

Empirical specification
To estimate differences in wealth across men and women in SSCs and DSCs, respectively, we use unconditional quantile regression (UQR).By estimating UQRs, we handle the skewness of wealth.We follow the approach suggested by Firpo et al. (2009) and use recentered influence function (RIF) UQRs to allow estimates to differ over the distribution.As a robustness check, we report in Appendix Table A4 the main results when using an inverse hyperbolic transformation, as suggested by Pence (2006).For ease of interpretation, we choose to present the results without the transformation in the article; however, the results are qualitatively similar.
where Y it is the wealth outcome of each individual i at a given time t.Ever in an SSC is an indicator variable equal to one for individuals who have ever been observed in a same-sex legal union (such as same-sex marriage or registered partnership); and c is a vector of regression coefficients for X it , where X it is a vector of time and individual demographic characteristics, in addition to age controls, time FEs and sibling order FEs.d is a vector of regression coefficients of sibling FEs in S i that captures unobserved factors that vary across sets of siblings, such as family background, family upbringing environment and shared family experiences.The error term e it in Equation ( 1) is assumed to be independent and identically distributed.Our coefficient of interest b captures the relative association between individuals in SSCs and those in DSCs by gender and wealth, as described above.By using sibling FEs, we address potential selection biases in our identification of sexual orientation, as well as confounding factors such as family background and values, including social origins, childhood geographical characteristics, parental educational attainment and labor market outcomes.Relying on legal unions for detecting sexual orientation may lead to a selection bias, disproportionately representing individuals in same-sex legal unions who feel safer disclosing their sexual minority status.This disclosure may be influenced by family values, potentially affecting other decisions and outcomes related to wealth accumulation.To account for this, we follow Murray (2002) and Rothblum et al. (2004), who used sibling techniques to compare individuals by sexual orientation and gender; a large body of literature has used sibling FEs to account, for example, for parental cognitive skills (Anger and Heineck, 2010;Gr€ onqvist et al., 2017), parental labor market outcomes (Bj€ orklund et al., 2010;Rothwell and Massey, 2015;Hilger, 2016;Liss et al., 2023), childhood geographical characteristics (Brattbakk and Wessel, 2013;Rothwell and Massey, 2015), school quality (Bedi and Edwards, 2002), parental wealth (Karagiannaki, 2017a), parental educational achievements (Humlum et al., 2019), parental involvement in schoolwork, parenting practices and maternal attitudes (Bj€ orklund et al., 2010) and the impact of inheritance on wealth (Karagiannaki, 2017b).As we compare siblings who share similar upbringings and family environments, the sibling approach enhances the causal interpretation of our findings, strengthening the validity and reliability of our results.
Moreover, as reported in Figure 2, individuals are more evenly distributed across ages in the sibling sample, regardless of union type and gender, but not in the full sample.By using the within-sibling approach, we thereby address this demographic asymmetry between individuals who have ever been in an SSC and those exclusively in DSCs.

Descriptive statistics
Table 2 presents the descriptive statistics of the sibling administrative data, providing an overview of the demographic and geographical characteristics of the sample.Specifically, we report sample averages for men ever in an SSC (Column 1), men exclusively in DSCs (Column 2), women ever in an SSC (Column 3) and women exclusively in DSCs (Column 4).The table includes information on age, sibling order, labor income, marital status and regional categories of the municipality of residence for our sample of individuals who had Downloaded from https://academic.oup.com/ser/advance-article/doi/10.1093/ser/mwae041/7695741 by guest on 20 June 2024 ever entered a legal union.Regional categories are designed to capture urban/rural differences.Notably, between 1995 and 2020, we had approximately 5300 sisters and 4400 brothers, approximately 2400 and 1900 of whom had ever entered same-sex unions, respectively.In the Appendix, we present various tables and figures describing the data in more detail: Appendix Table A1 contains detailed descriptions of each variable; Appendix Figure A2 the number of individuals in the sibling sample by gender and sexual orientation (1995-2020); Appendix Figure A3 shows the number of individuals in legal union in the sibling sample by gender and sexual orientation (1995-2020); and Appendix Figure A4 shows the sibling order distribution of the sibling sample in 2003.
The demographic and geographical characteristics of the siblings displayed in Table 2 validate the findings obtained utilizing the full population from previous economic research ( Andersson et al., 2006;Ahmed et al., 2011;Ald� en et al., 2015).Specifically, our analysis reveals that women and men who have ever been in an SSC tend to be younger, more highly educated and more likely to reside in urban areas than their sibling counterparts.Table 2 shows the labor income penalty for men who have ever been in an SSC and the labor income premium for women who have ever been in an SSC.Moreover, Table 2 suggests that both men and women in SSCs have lower levels of real estate wealth, real estate ownership rates and total debt than their siblings exclusively in DSCs.However, on average, men who have ever been in an SSC appear to not have significant differences in total assets nor in net wealth compared to their brothers exclusively in DSCs, while the raw data indicate that women who have ever been in an SSC have lower total assets and net wealth than their sisters exclusively in DSCs.Siblings of the same gender but with different sexual orientations differ in terms of their relative position in the wealth distribution, especially for the higher deciles, as illustrated in Figure 3.As expected, the figure shows that brothers in our sample are overrepresented in the top deciles of the wealth distribution compared with sisters.On average, men and women who have ever been in an SSC have less or as much wealth as their siblings exclusively in DSCs at every decile of the wealth distribution.

Sexual orientation and wealth outcomes
Using UQR, we start by estimating how having ever been in an SSC is related to net wealth and its subcomponents.The estimations are performed separately for male and female siblings at various percentiles in the respective wealth distribution.Table 3 presents the estimates of having ever been in an SSC, estimating Equation (1) by gender with age controls and sibling, sibling order and year FEs for net wealth.Panel A suggests that the effect of ever having been in an SSC varies considerably over the net wealth distribution for men.At the median (P50), there is a penalty for men having ever been in an SSC, while in the top two deciles, there is a sizeable premium compared to their brothers exclusively in a DSC.As reported in Panel B, the trend is different for women, where the net wealth penalty for women who have ever been in an SSC becomes insignificant in the top decile (P90); otherwise, the penalty is consistently present in the female net wealth distribution.For instance, at the median (P50), women who have ever been in an SSC have 22 316 SEK (approximately e2500), less net wealth than their sisters exclusively in a DSC.Estimation results without age controls are presented in Appendix Table A7 and indicate that while age does significantly impact wealth accumulation, the overall findings in Table 3 remain consistent.Figure 4 reports the estimated wealth disparities between siblings who have ever participated in an SSC and their counterparts who were exclusively involved in DSCs for four wealth components (i.e.real estate wealth, financial wealth, student debt and total debt).There is considerable variation in the differences across sexual minority status and gender in the subcomponents of wealth.
First, it is worth noting that up to half of the sample has no real estate wealth, financial wealth, total debt or student debt.At the top, men who have ever been in an SSC have more real estate and financial wealth, as well as more debt (in both total and student debt), than their male siblings exclusively in a DSC.Women who have ever been in an SSC have worse outcomes because they have less real estate wealth and more student debt than their sisters exclusively in a DSC.At the same time, women in an SSC have less total debt than women in a DSC.
Consistent with the findings of previous research and our descriptive statistics reported in Table 2, men and women who have ever been in an SSC tend to have higher levels of educational attainment than their heterosexual counterparts (Andersson et al., 2006;Ahmed et al., 2011;Badgett et al., 2021).In line with this, we find that individuals who have ever been in an SSC, both men and women, accumulate significantly more student debt than their heterosexual siblings.A higher level of education attained by sexual minority men and women is likely associated with the accumulation of greater student debt for this group.Our results suggest that this disparity in student debt plays a significant role in the wealth gap across sexual orientation and gender lines.

Sexual orientation, wealth and mechanisms
To explore possible channels explaining the net wealth differentials within the sibling sample, we re-estimate Equation ( 2) to test two plausible mechanisms: education and the urban/ rural channels.An individual is living in an urban area if the individual is residing in a municipality that has less than 50% of their population in rural areas (see Appendix Figure A5 for the regional category of every municipality in Sweden).Years of education is a measure of an individual's educational attainment and reflects the number of academic years an individual has achieved in structured learning environments up to that given year.In the Swedish context, secondary education takes 12 years, and the completion of a bachelor's degree generally an additional 3 years (in total 15 years of education).Figure 5 shows the marginal effects of urban areas and years of schooling on the net wealth gap. Figure 6 displays the marginal effects of years of schooling on the net wealth gap for siblings who have ever been in an SSC in urban and rural areas (Appendix Tables A2 and A3 report the UQR and the marginal effects for P99, P95 and all deciles.).
For women who have ever been in an SSC, the marginal effects of urban residency on net wealth are negatively significant in the lowest quantiles and positively significant in the highest   quantiles (Figure 5, left panel).The lower marginal effect for the lowest quantiles indicates that at the lower end of the distribution, being an urban resident has a relatively negative effect on women who have ever been in an SSC.The higher marginal effect for the highest quantile suggests that at the upper end of the distribution, being an urban resident has a significant positive impact on the net wealth of women who have ever been in an SSC.For men who have ever been in an SSC, the urban marginal effect is either positive or non-significant throughout the distribution but significantly greater at the upper end of the distribution than for women.The marginal effect of years of schooling on the net wealth gap provides insights into how one year of schooling change influences the net wealth gap (Figure 5, right panel).In our case, we observe that an increase in years of schooling is associated with a negative net wealth gap for women who have even been in an SSC relative to their exclusively being in DSC sisters throughout the net wealth distribution.
Conversely, while an increase in years of schooling is associated with a negative net wealth gap at the lowest quantile for men, a positive impact on the net wealth gap is observed at the highest quantile.This finding implies that at the highest quantile, additional years of schooling are positively associated with net wealth for men who have ever been in an SSC relative to their brothers who are exclusively in a DSC.A possible explanation is that higher levels of education might be associated with increased opportunities, betterpaying jobs or other factors that contribute to greater wealth accumulation for men who have ever been in an SSC than for their brothers exclusively in a DSC in the highest quantile.For women who have ever been in an SSC, however, the return on education is, on average, lower than for women exclusively in a DSC, and additional years of schooling do not change the wealth penalty by sexual orientation.Now, we focus on the points in the net wealth distribution, P20, P50 and P90; see Table 4 (Appendix Tables A2 and A3 report the UQR for all deciles.).
For both men and women at P20, an additional year of schooling is negatively associated with wealth, while the impact of education on the wealth gap for individuals who have ever been in an SSC compared to their siblings exclusively in DSCs varies between genders; that is, there is no significant effect for men who have ever been in an SSC but negative effects for women who have ever been in an SSC compared to their siblings exclusively in a DSC.This negative impact of education disappears only at P90, where men, in general, enjoy a positive return on education, especially men who have ever been in an SSC.However, women who have ever been in an SSC with an additional year of schooling have less wealth than their sisters exclusively in a DSC.
To disentangle the effects of education and urban location on net wealth, we divide our sample based on whether the individuals who have ever been in an SSC reside in urban or rural areas.
For both genders, Table 4 and Figure 6 suggest that the above results are driven by SSC siblings living in urban rather than rural areas (as reported in Table 2, most men ever in SSC reside in urban areas).Moreover, Table 4 and Figure 6 suggest that the positive return on years of schooling for men is driven by brothers being the siblings who have ever been in an SSC living in urban rather than rural areas.

Robustness
In this subsection, we re-estimate Equation (1) for the entire population, that is, including singletons and siblings (without any sibling FEs), and for the sibling population, omitting Downloaded from https://academic.oup.com/ser/advance-article/doi/10.1093/ser/mwae041/7695741 by guest on 20 June 2024 the sibling FEs.Comparing these new results with our main results in Table 3 enables us to examine the external validity of our main results.
According to both the full population (Table 5) and the sibling population analyses (Figure 7), the estimated net wealth gap for women who have ever been in an SSC is consistently and significantly negative at P80 and below.At the top deciles, the results are significantly negative for the full population and non-significant for the sibling population (omitting the sibling FEs).These results suggest that, irrespective of whether considering the broader population or a more controlled sample of siblings, women in an SSC tend to have lower accumulated wealth than their counterparts in a DSC.However, the magnitude of the estimate for women who have ever been in an SSC is greater in the full sample than in the within-siblings sample, conceivably due to the larger average age difference between women in SSCs and those in DSCs in the full sample.Conditioning on age (as we do throughout) in the UQR is insufficient to capture wealth accumulation over the life cycle,  leading to an overestimation in the full sample of the negative effects on wealth for women who have ever been in an SSC.
When comparing the results between the full population (Table 5) and the subset comprising siblings with siblings FEs (Table 3), we observe that within the top decile (P70 and above), men who have ever been in an SSC exhibit a positive net wealth premium.This indicates that in this upper wealth distribution range, men who have ever been in an SSC tend to accumulate more wealth than their counterparts exclusively in a DSC.Conversely, at the bottom decile (P50 and lower), there is a negative net wealth gap, implying that men who have ever been in an SSC have lower wealth accumulation than men in DSCs.Notably, both the full population and sibling population estimations reveal significantly negative values for the net wealth gap at P50 and lower, indicating a consistent trend across these analytical approaches.
We plot the estimates of having ever been in an SSC with sibling FEs (left panel) and without sibling FEs (right panel) in Figure 7 to examine the importance of family fixed characteristics; the full set of estimates is reported in Appendix Table A6.The estimates of ever being in an SSC are larger in the right-hand panel when excluding sibling FEs, as shared family characteristics are no longer accounted for.A comparison of the panels in Figure 7 shows that not controlling for family unobservables leads to an overestimation of the sexual orientation wealth gap by gender.
One asymmetry between individuals who have ever been in an SSC and those exclusively in DSCs is that it became possible to enter a same-sex legal union only a few years before the start of our period of analysis.Therefore, before 1999, there were many more DSCs than SSCs.However, as shown in Figure 8, when controlling for being in a legal union at time t, the results for the net wealth gap are consistent with our main results, although somewhat smaller because individuals in an SSC are, on average, younger than individuals in a DSC.In addition, in Appendix Table A5, we report the UQR estimate of the effect of legal union on net wealth by gender and sexual orientation for siblings (in SEK), and in Appendix Table A8, we report mean age at different wealth quantiles for individuals in an SSC and those in a DSC in 2003 by legal union status, showing that both men and women in an SSC are considerably younger than their siblings in a DSC.

Discussion and conclusion
Our study provides the first comprehensive evidence of the relationship between sexual orientation and wealth by gender.As Sweden has been at the forefront of progressive policies targeting sexual minorities, our findings are indicative of the relationship between wealth and sexual orientation by gender within the context of a highly progressive society.Our study benefits from the use of detailed and high-quality, third-party-reported, register-based individual wealth data.Moreover, we are able to estimate sibling FEs, which enables us to compare sexual minorities with similar heterosexual individuals.This approach provides a more rigorous analysis, controlling for potential confounding factors that may arise from family background and norms and other environmental factors during upbringing, strengthening the causal interpretations of our results.
We find that men in SSCs possess, on average, 19% more wealth than their brothers who have been exclusively in DSCs in the top decile of the net wealth distribution and, on average, 55% less than their brothers in P50.Women who have ever been in an SSC tend to have less wealth than their heterosexual sister(s) over the entire net wealth distribution, varying between 5% and 179% less, except for the top decile.This is unexpected, as the literature on labor market outcomes shows that sexual minority men experience an income penalty, while sexual minority women enjoy an income premium relative to their heterosexual counterparts (Andersson et al., 2006;Ahmed et al., 2011;Ald� en et al., 2015).Our findings suggest that further research is needed to understand whether wealth accumulation differs according to gender and sexual orientation or whether these differences are the result of the interplay of variation in wealth distribution together with age and historical barriers to entering into legal unions.
Regarding the different wealth components, SSC women and SSC men have less real estate wealth, less total debt, and more student debt between P50 and P70.However, in P90, these groups have more financial wealth than their siblings exclusively in a DSC.Except for the very high real estate wealth distribution for men, our results are in line with those of previous studies suggesting that sexual minority individuals face real estate credit market discrimination, leading to lower rates of mortgage acceptance and real estate ownership (Dillbary and Edwards, 2019;Sun and Gao, 2019).Additionally, individuals who have ever been in an SSC have somewhat more financial assets.One possible explanation for this phenomenon, in line with discrimination theory (Dillbary and Edwards, 2019;Sun and Gao, 2019), is that sexual minority individuals are significantly more likely to be excluded from the credit and real estate markets and, instead, invest in financial assets.
When investigating the underlying factors contributing to wealth disparities within the sibling sample, our analysis identifies two significant mechanisms: education and the rural/ urban divide.Specifically, SSC individuals who have ever been in an SSC are more likely to reside in urban and more tolerant areas, exposing them to unique economic urban environments; in particular, they present distinct advantages for wealth accumulation, especially for men and women who have ever been in an SSC, as evidenced by our findings.
In line with the previous literature, our results suggest that the residential concentration of sexual minority individuals in large metropolitan areas could increase human capital, well-being levels and wealth disparities through sexual orientation (Black et al., 2000(Black et al., , 2002;;Gates and Ost, 2004;Denier and Waite, 2018;Baumle et al., 2020).After moving to more progressive urban regions (Badgett et al., 2021), sexual minority individuals encounter better human capital, amenities, income, wealth and economic opportunities (Black et al., 2002;Gates and Ost, 2004;Florida et al., 2008;Florida, 2019).The geographic channel, therefore, emerges as a crucial determinant in shaping the financial outcomes of sexual minority individuals.
Moreover, we know that sexual minority individuals have higher levels of education (Mittleman, 2022).Our investigation highlights the pivotal role of education in contributing to the observed net wealth disparities within the sexual minority population.Notably, our findings reveal nuanced dynamics, indicating that the net wealth return on additional years of schooling varies across percentiles and gender.Specifically, at P50 and P90, a positive return on years of schooling is observed for sexual minority men, while it is negative for sexual minority women.Interestingly, conditioning on residing in a rural area reverses the positive return on education for men in an SSC, transforming it into a negative impact on net wealth.
One possible interpretation of the observed impact on the wealth premium for men, when controlling for education, is rooted in the progressive nature of urban areas, which tend to offer more economic opportunities for highly skilled workers, irrespective of gender.However, it is crucial to note that higher education levels of sexual minority individuals in more expensive urban settings contribute to increased student debts.For urban men who have ever been in an SSC, the positive urban marginal effects outweigh student debts, resulting in a positive return on schooling.Conversely, women who have ever been in an SSC experience smaller positive urban marginal effects that fail to compensate for greater student debt, leading to a negative return on schooling.These results underscore the need to consider the complex interplay between individuals, education and geographical factors in understanding the relationship between sexual orientation and wealth.
Despite the strengths of our study, there are certain limitations, primarily related to the data available for analysis.Although the use of third-party reported administrative data allows for a large sample size and high-quality data on sexual minority individuals, our study is unable to identify individuals who have always been legally single and, therefore, may not capture the experiences of all sexual minority individuals.Additionally, previous research has suggested that bisexual individuals are more likely to enter different-sex relationships, which means that our data may not be able to fully capture the experiences of this group.Therefore, future research should continue to investigate these groups.Moreover, the collection of wealth data was discontinued in Sweden in 2007, making it impossible for us to extend the analysis to the present day.While there are ongoing efforts in Sweden to reinstate the collection of wealth data for statistical analysis and research, the wealth register used herein, albeit somewhat dated, remains a very valuable source for analyzing wealth inequality in an entire population.
While our study provides a significant contribution to the literature on wealth disparities associated with sexual orientation and gender, importantly, our findings focus on Sweden.Future research should seek to extend our findings by examining similar associations in different legal and cultural contexts.While our study provides insight into several potential mechanisms that may help explain the observed phenomena, further research exploring additional mechanisms is necessary to fully grasp the complexities of this issue across diverse settings.

Figure 1
Figure 1 Number of individuals in legal unions by union type and year.

Figure 2
Figure 2 Age histogram of the full population (Panel A) and in the sibling sample (Panel B)-2003.

Figure 3
Figure 3 Net wealth distribution in the sibling sample in 2003, by decile.

Figure 4
Figure 4 The sexual orientation wealth gaps by gender for siblings (in SEK) with different net wealth components.

Figure 5
Figure 5 The urban (left panel) and years of schooling (right panel) marginal effects on the net wealth gap by sexual orientation for siblings (in SEK).

Figure 6
Figure 6 Years of schooling have marginal effects on the net wealth gap (in SEK) for siblings ever in an SSC in urban areas (left panel) and rural areas (right panel).

Figure 7
Figure 7 The sexual orientation wealth gaps by gender for siblings (in SEK) with sibling FEs (left panel) and without sibling FEs (right panel).

Figure 8
Figure 8 The sexual orientation wealth gaps (in SEK) by gender for siblings (left panel) and with legal union control (right panel).
Dependent variable: Net wealth in SEK.Robust standard errors are in parentheses.Author calculations from Sweden's population register.Dependent variable: Net wealth in SEK.Robust standard errors are in parentheses.Author calculations from Sweden's population register.Dependent variable: Net wealth in SEK.Robust standard errors are in parentheses.Author calculations from Sweden's population register.

Figure A1
Figure A1 Age distribution of the first legal union across the full Swedish population (1995-2020).

Figure A2
Figure A2 Number of individuals in the sibling sample.

Figure A4
Figure A4 Sibling order distribution of the sibling sample in 2003.

Figure A5
Figure A5Regional categories of each municipality in Sweden.

Table 1
Wealth distribution, age and education by gender and union type in 2003 Downloaded from https://academic.oup.com/ser/advance-article/doi/10.1093/ser/mwae041/7695741 by guest on 20 June 2024 Our empirical specification is as follows: Notes: Table1shows the age, education level, the wealth distribution quantiles by gender and union types for the adult and sibling population in Sweden in 2003.8E.Dujeancourt et al.

Table 2
Summary statistics of the siblings Downloaded from https://academic.oup.com/ser/advance-article/doi/10.1093/ser/mwae041/7695741 by guest on 20 June 2024 Riksbank, 2023)ssets, debts and income are denominated in real SEK with base year 1999 (SCB, 2023).The average SEK/EUR exchange rate between 1999 and 2007 was 1 EUR ¼ 9.079 SEK (SverigesRiksbank, 2023).Siblings have identical biological parents, are of the same gender, and have at least one sibling ever in an SSC.��� P < 0.01, �� P < 0.05, � P < 0.1 indicate the statistical significance of the difference in means with gender by union type (Columns 1 and 2 for men; Columns 3 and 4 for women).10E.Dujeancourt et al.

Table 3
UQR estimates of net wealth by gender and sexual orientation for siblings (in SEK) Notes: Dependent variable: Net wealth in SEK.Robust standard errors are in parentheses.Author calculations from Sweden's population register.��� P < 0.01,

Table 5
UQR estimates of net wealth by gender and sexual orientation for the full population (in SEK)

Table A1 Continued
Table A4Unconditional quantile inverse hyperbolic sine transformation estimates of net wealth by gender and sexual orientation for siblings (in SEK) Robust standard errors are in parentheses.Author calculations from Sweden's population register.