Background Socioeconomic status (SES) is a fundamental contributor to health; however, limited research has examined sexual orientation differences in SES.
Methods 2008–2009 data from 14 051 participants (ages 24–32 years) in the US-based, representative, National Longitudinal Study of Adolescent to Adult Health were analysed using multivariable regressions that adjusted for age, race-ethnicity, childhood SES, urbanicity and Census region, separately for females and males. Modification by racial minority status (black or Latino vs white, non-Hispanic) was also explored.
Results Among females, sexual minorities (SM) (10.5% of females) were less likely to graduate college, and were more likely to be unemployed, poor/near poor, to receive public assistance and to report economic hardship and lower social status than heterosexuals. Adjusting for education attenuated many of these differences. Among males, SM (4.2% of males) were more likely than heterosexuals to be college graduates; however, they also had lower personal incomes. Lower rates of homeownership were observed among SM, particularly racial minority SM females. For males, household poverty patterns differed by race-ethnicity: among racial minority males, SM were more likely than heterosexuals to be living at >400% federal poverty level), whereas the pattern was reversed among whites.
Conclusions Sexual minorities, especially females, are of lower SES than their heterosexual counterparts. SES should be considered a potential mediator of SM stigma on health. Studies of public policies that may produce, as well as mitigate, observed SES inequities, are warranted.
- social inequalities
- health inequalities
- social science
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Health inequalities by sexual orientation have been widely documented in every domain of health,1–4 including: violence victimisation,5–9 tobacco use,10 11 suicidality,12–15 poor mental health16–19 and healthcare barriers.20 HIV/AIDS has exacted a prolonged toll on gay and bisexual men.21 22 Obesity23 24 and disability,2 25 have, more recently, emerged as lesbian health concerns.
Socioeconomic status (SES) is a fundamental contributor to health and disease across the life course,26–28 and varies by sexual orientation; however, SES is often treated as a statistical control and is rarely discussed as a potential mediator of health inequities experienced by sexual minorities (SM). Inadequate economic resources are associated with poor health28–32 through both material and psychosocial pathways that increase exposure to hazards and decrease exposure to health-promoting resources.26 33 Consistent with a socioeconomic gradient in health, several,25 34 35 but not all,2 population-based studies report higher rates of poverty among SM compared with heterosexuals. Yet, these findings vary by sex,25 34 sexual orientation,25 34 35 selection of statistical controls34 and place.2 25 35 For instance, nationally, higher poverty rates were found among female same-sex couples than among married different-sex couples, whereas, among males, poverty rates were lower among same-sex couples.34 However, after adjusting for education, employment and demographic characteristics, poverty rates were higher for same-sex male couples compared with married different-sex couples.34
In addition to variability in findings that used income-based measures of economic status, peer-reviewed research has yet to examine sexual orientation differences in assets, financial hardship and subjective social status—aspects of SES that have been linked to health in general population samples.28 30 36–38 Understanding the breadth and nature of sexual orientation differences in SES is essential to reducing health inequities, particularly as the size of the SM population grows39 and ages. The current study addresses these gaps in knowledge and examines a comprehensive array of SES indicators across sexual orientation groups, separately by sex, in a large, population-based sample.
The National Longitudinal Study of Adolescent to Adult Health (Add Health) is a nationally representative, longitudinal study of US adolescents initiated in 1994 and conducted by the Carolina Population Center at the University of North Carolina. In the 1994–1995 academic year, a total of 20 745 adolescents enrolled in grades 7–12 completed baseline in-home surveys. Add Health, whose methods have been well-described elsewhere,40 is currently in the field with the wave V survey. The current study focused on outcomes measured in the young adulthood/wave IV survey, conducted in 2008–2009 when respondents were aged 24–34 years. Eligibility for the current cross-sectional study was limited to those who completed baseline and wave IV surveys (n=15 701; 80.3% of original baseline sample) and for whom a wave IV sampling survey weight was available (n=14 800). Missingness due to the lack of a Wave I sampling weight was administrative in nature,41 and thus, was ignorable42 in relation to the analyses presented in this manuscript. The final analytic sample included 14 051 respondents (93.7% weighted, of those eligible) who provided data on sexual orientation and covariates.
Sexual orientation identity was embedded in the computer-assisted self-interviewing portion of the interview which has been shown to increase disclosure of ‘sensitive’ subject matter.43 Respondents who selected bisexual, mostly homosexual or 100% homosexual options as their sexual orientation identity at wave IV were classified as SM while those who selected 100% heterosexual were classified as heterosexual. Self-reported mostly heterosexuals (n=1368) were classified as sexual minorities if they reported one or more lifetime same-sex sexual partners (n=528); otherwise, they were grouped with heterosexuals (n=840).
Prior to grouping all SM, we conducted a sensitivity analysis to determine whether bisexually identified individuals should be grouped with other SM (vs treated separately) given that bisexuals in Washington state and Massachusetts25 35 were found to have the lowest SES of all sexual orientation groups. In multivariable regressions, we observed that the pattern (direction and magnitude of associations between sexual orientation and SES indicators) was similar in models that included and excluded bisexuals (n=214). Consequently, we created one SM group.
Respondents were classified as male or female based on their responses to a wave I question, "What is your sex?"
Wave IV SES
Educational attainment was parameterised as <high school (HS)/graduate equivalence degree (GED), HS/GED, some college or vocational education and ≥bachelor’s degree. Wave IV employment status was coded as currently employed (10 or more hours per week for pay), unemployed, homemaker, student and other (not employed due to disability, temporary parental leave, activity military service or incarceration). Personal income in the prior year, before taxes and deductions and including non-legal sources, was categorised as <US$10 000, US$10 000– US$24 999, US$25 000–US$49 999 and ≥US$50 000. Respondent-reported annual household income and size were used to create an ordinal measure of percentage poverty. Annual household income, also collected categorically, was recoded to the mid-point for each income range or, for those who selected the highest category (≥US$150 000), to the 95% percentile of 2007 annual family income (US$197 216).44 Recoded income was divided by size-specific poverty thresholds45 to obtain percentage of the federal poverty level (FPL) (ie, the ’income-to-needs ratio').46
Receipt of public assistance in adulthood was indicated if the respondent, or anyone in their household, had received public assistance, welfare payments or food stamps since their last interview in 1995 (wave II) or 2001–2002 (wave III). Economic hardship in the prior 12 months was indicated by endorsement of any of six indicators, created for Add Health, unless otherwise noted. These were: went without phone service, did not pay full amount of the rent or mortgage, did not pay full gas, electricity or oil bill, evicted from house or apartment, had gas/electricity/oil utility service shut off or were worried whether food would run out before being able to buy more because they did not have enough money.47 Current homeownership was indicated by a yes to the question, "Is your house, apartment, or residence owned or being bought by (YOU AND/OR YOUR SPOUSE/PARTNER)?” The MacArthur Scale of Subjective Social Status SES Ladder48 was used to assess subjective social status. Respondents were asked to indicate where they fell on a ladder from 1 to 10 (1 being ‘the people who have the least money and education, and the least respected jobs or no job’) relative to other people in the USA.
A number of self-reported sociodemographic characteristics, associated with both sexual orientation and SES, were treated as potential confounders. These included: age (24–27, 28–29, 30–34 years) and wave I race-ethnicity, which was coded hierarchically as any Hispanic ethnicity, black, Asian or Pacific Islander or American Indian or self-reported ‘other’ race and white. Parental education at wave I was defined as the highest attainment obtained by a parent/guardian (less than a HS diploma, HS or GED, some college, vocational school or post-HS training, >bachelor’s degree) as reported by the respondent or the parent/guardian. Receipt of public assistance in childhood was indicated if anyone in the household received ‘public assistance, welfare payments or food stamps’ before the respondent was 18. Data were collected at wave III or at wave IV if wave III data were missing. Wave IV Census region (Northeast, Midwest, South, West and wave IV urbanicity were based on the respondent’s Census tract. Census tracts with density below 1000 people/square mile, as per 2009 American Community Survey 5-year estimates, were characterised as rural49; all others were categorised as urban.
Descriptive analyses were conducted to compare the distribution of SES indicators and covariates across sexual orientation groups separately by sex (table 1). Sex-stratified multinomial and binary logistic regression models were fit for each SES indicator to generate relative risk ratios (probability ratios) or ORs, respectively, using the following model-building approach: a) crude (model 1 as shown in table 2); b) adjusted for covariates (age, race-ethnicity, highest parental education, receipt of public assistance <age 18 years, urbanicity and Census region (model 2), c) adjusted for education plus all covariates (model 3) and d) adjusted for employment status, education, plus all covariates (model 4). This model-building approach allowed us to examine associations between sexual orientation and SES, with and without adjustment for education and employment status. In order to provide information about the SES distribution of each sexual orientation group, adjusting for potential confounders, but not accounting for factors on the casual pathway (ie, education and employment status), predicted probabilities (categorical outcomes) and average values (continuous outcome) for each SES indicator were computed separately by sexual orientation using the margins command in STATA, following model 2 and reported in table 3.
In order to explore potential effect modification by racial minority status (operationalised as black or Latino) versus the dominant group (white, non-Hispanic), SM by racial minority interaction terms were added to model 2 regressions. Because the ‘other’ racial-ethnic group was small and heterogeneous in terms of racial-ethnic identity, SES and group histories of racism, it was excluded from these analyses. The presence of a statistically significant interaction term at an alpha of 0.10 was used to determine the presence of possible effect modification (see online supplementary table A). Predicted probabilities were computed separately by sex, sexual orientation and racial minority/majority status and graphed for SES indicators where the association between sexual orientation and SES appeared to vary across racial minority/majority status. All analyses were conducted in STATA V.14,50 incorporating Add Health sampling weights and adjusting for the complex sampling design.
Characteristics of the analytic sample are presented in table 1. Most (92.7% weighted) of the respondents aged 24–34 years were heterosexual; however, 7.3% of respondents were categorised as SM because they reported bisexual, mostly homosexual or 100% homosexual identities or reported one or more lifetime same-sex sexual partners (if mostly heterosexual). A higher proportion of females (10.5%, n=761) were classified as SM than males (4.2%, n=295).
Among females, SM were over-represented among those who did not complete an HS or GED and were under-represented among those who completed ≥bachelor’s degree compared with heterosexual females (table 1). Most females, across sexual orientation groups, were employed; however, SM females were somewhat under-represented among the employed and over-represented among the unemployed. SM females were slightly over-represented in the group reporting <US$25 000 in personal annual income, as well as in the near poor (100%–199% FPL) and highest (≥400% FPL) economic status groups. SM females were also more likely to report receipt of public assistance since the last interview, as well as economic hardship in the prior year, compared with heterosexual peers. They were also less likely to be homeowners and reported lower mean subjective social status scores.
After adjusting for covariates, the risk of completing ≤bachelor’s degree was significantly higher for SM females relative to heterosexual peers (table 2, model 2). In fact, the risk of not completing HS was three and a half times greater (relative risk ratio (RRR), 3.5, 95% CI 2.3 to 5.4), and the risk of completing a HS/GED or some college was twice as great (RRR 2.1, 95% CI 1.4 to 3.2; RRR 2.1, 95% CI 1.6 to 2.9, respectively). SM females were also more likely to be unemployed (RRR 2.2, 95% CI 1.5 to 3.3), to earn US$10 000–US$25 000 vs ≥US$50 000 in the prior year (RRR 1.5, 95% 1.1 to 2.1), and to be near poor (100%–199% FPL; RRR 1.6, 95% CI 1.1 to 2.2) versus at ≥400% FPL. The odds of reporting public assistance since the last interview (OR 1.5, 95% CI 1.2 to 1.8) and any economic hardship in the prior year (OR 1.7, 95% CI 1.4 to 2.1) were elevated, while the odds of homeownership (OR 0.6, 95% CI 0.4 to 0.7) were reduced among minority versus heterosexual females. Subjective social status scores were an average of 0.4 points (95% CI −0.5 to −0.2) lower among SM females. Adjusting for respondent education (model 3) attenuated employment, receipt of public assistance and income-based indicators of SES; however, unemployment, homeownership, economic hardship and subjective social status remained statistically significantly different between SM and heterosexual females. Further adjustment for employment status (model 4) did not alter the pattern of results.
Among females, the association between sexual orientation and SES varied across racial minority versus majority groups for homeownership (F=3.80, df(1, 128), p=0.053) (figure 1). Differences in rates of homeownership by sexual orientation appeared larger among whites (38.5% SM vs 53.2% heterosexual) than among racial minorities (24.7% SM vs 30.4% heterosexual). Notably, rates of homeownership were lower among racial minorities and were the lowest among racial minority SM females.
A larger proportion of SM males completed ≥bachelor’s degree compared with heterosexual males (table 1). The vast majority of males (approximately 85%) were employed across sexual orientation groups. SM males were over-represented at lower levels of personal income, but did not statistically significantly differ on the household-size adjusted poverty-to-income needs ratio. SM males were less likely to be homeowners than their heterosexual peers.
After adjusting for covariates (table 2, model 2), the risk of having an HS/GED compared with ≥bachelor’s degree was significantly lower (RRR 0.4, 95% CI 0.2 to 0.6) for SM males than heterosexual males. SM males were also more likely to earn <US$10 000 (RRR 2.2, 95% CI 1.2 to 4.2) and US$10 000–US$25 000 (RRR 2.1, 95% CI 1.2 to 3.7) vs ≥US$50 000 in the prior year than heterosexual males. The odds of homeownership (OR 0.4, 95% CI 0.3 to 0.6) were considerably lower among SM males. Adjusting for respondent education (model 3) magnified these inequities, indicating that, given high levels of education, SM males, on average, have fewer economic resources than expected and are at increased risk of economic hardship (OR 1.6, 95% CI 1.1 to 2.3.) Adjustment for employment status (model 4) did not alter the pattern of results.
Among males, the association between sexual orientation and SES varied across racial minority versus majority groups for employment status (F=132.84, df(4, 128), p<0.001) and household poverty (F=2.43, df(4, 128), p=0.0514) (figure 2). Since most males, across sexual orientation and racial minority/majority groups, were employed (81.3%–85.0%), the other employment status categories included relatively few respondents and thus CIs around these estimates were quite wide. For instance, the predicted probability of unemployment was 4.0% (95% CI 0.5 to 7.5) for SM white males, 5.9% (95% CI 4.6 to 7.3) for heterosexual white males, 7.5% (95% CI −0.8 to 15.7) for racial minority SM males and 9.6% (95% CI 7.3 to 11.9) for racial minority heterosexual males. Given the instability of these estimates, and the lack of a clear pattern to report, no figure is included for employment. In contrast, the pattern observed for household poverty was clearer. The association between sexual orientation and household poverty was reversed across race, such that SM racial minority men were more likely to be living at ≥400% FPL than racial minority heterosexual men (48.0% vs 32.1%, respectively), whereas SM white men were less likely to be in the highest economic status group than their heterosexual white male counterparts (38.3% vs 46.7%, respectively).
Socioeconomic inequities were observed among SM, particularly females, in the population-based Add Health sample. SM females were less likely to complete ≥bachelor’s degree, were more likely to be unemployed, to be near poor, to receive public assistance and to report economic hardship. They also reported lower subjective social status, which is unsurprising given that their objective SES was lower than that of heterosexual women and of SM men in this study.
Many of the observed economic inequities among women appeared to be related to differences in educational attainment. Economic inequities were attenuated after adjusting for education—suggesting that promoting the achievement of SM girls and young women may serve to reduce economic inequalities—regardless of the temporal ordering between educational completion and the expression of SM status. Proximal or ‘midstream’ factors that may underlie this gap include sexual victimisation,6 unplanned pregnancy51 52 and differential discipline in secondary schools,53 all of which are more common among SM women, and all of which are inversely associated with education.
Fewer significant sexual orientation differences in economic status emerged among males, which may be due to higher levels of education among minority males. In contrast to the pattern observed among females, SM males were more likely to complete college. This was an unexpected finding given that SM men report higher rates of school harassment than their heterosexual peers.54 55 One potential explanation for this pattern may include an investment in academic achievement among SM males as a way to garner positive attention.56 However, SM males were more likely to report lower personal incomes, and, after accounting for higher levels of education, were more likely to report economic hardship in the previous year, than their heterosexual counterparts. This pattern, observed previously in Add Health,57 and as reported in a recent meta-analysis,58 suggests that SM males experience wage discrimination.
Given the relationship between household composition and size-adjusted household economic status, post hoc descriptive analyses of household composition were conducted. SM females were more likely to live with a same-sex romantic partner (9% vs 0%) or with others (eg, relatives, roommates) (33.0% vs 27.1%) than with a different-sex partner (49.3% vs 63.8%) and were as likely to live alone (8.7% and 9.2%, respectively) as heterosexual women. A large, but somewhat smaller (55.9% vs 62.6%) proportion of SM females were living with a son/daughter under the age of 18 as compared with heterosexual women. These data suggest that lower personal incomes among SM women are the likely driver of their over-representation among the near-poor rather than differences in household composition.
Among men, SM men were more likely to live with a same-sex partner (15.4% vs 0%) or to live alone (22.6% vs 12.6%) or with others (42.5% vs 30.0%) versus with a different-sex partner (19.5% vs 57.4%) than heterosexual peers. SM males were also far less likely to report living with a minor son/daughter (11.1% vs 41.1%) than their heterosexual counterparts. These data suggest that a lower likelihood of a (lower59) female wage earner and a child in the household, as reflected in smaller average household size, among SM males may help to explain why personal income inequities were not sustained across household economic status.
Although examining determinants of SES was beyond the scope of the present study, a social determinant of health framework, based on the Conceptual Framework for Action on the Social Determinants of Health60 (figure 3), was used to guide our reflections about putative causes of observed SES patterns and their impact on health. Importantly, in this framework, norms and values that privilege the dominant group (heterosexuals) and stigmatise others (sexual minorities) shape living and working conditions, including risk of sexual assault, access to health services and the presence of children in the household. Daily conditions are themselves influenced by governmental and institutional (eg, school discipline) policy.
Working through potential contributors to lower rates of homeownership among SM, as an illustrative example, we consider upstream determinants of material resources (both savings and income) and access to loans. Employment discrimination by sexual orientation is prohibited in only 22 states,61 is more commonly experienced by SM62 and may contribute directly to economic status through earnings (joblessness, underemployment), as well as, indirectly, by limiting access to employer-provided health insurance.63 Same-sex couples were not granted the right to marry across the USA until 26 June 201564; marriage facilitates access to mortgage loans,65 as well as health insurance coverage.66 Medical expenses related to lack of insurance or poor coverage impact savings and are significant contributors to bankruptcy.67 Strained parental relationships7 68 69 may further reduce access to material support (eg, housing,70 college tuition support, health insurance coverage, loans and gifts, loan cosignature) for SM. Intergenerational transfers are estimated to account for approximately 20% of personal wealth.71 Lastly, a preference or need to live in more tolerant (eg, those with local non-discrimination protections), but expensive urban areas72 73 may also impact economic resources and rates of homeownership.
Although an intersectional analysis that considers racial inequality as an important determinant of population patterns of SES was beyond the scope of the current paper, we did explore whether observed sexual orientation and SES patterns differed between racial minorities (black and Latino/as) and the majority (whites) separately for females and males. Patterns differed for 3 out of 16 SES indicators. Among women, sexual orientation inequities in homeownership were more pronounced for whites than racial minorities. Rates of homeownership were the lowest for SM racial minority women and highest for heterosexual white women. Among men, racial minority men were more likely to be in the highest household economic status group than were racial minority heterosexual men, whereas white SM men were less likely to be in the highest household economic status group compared with white heterosexual men. These patterns should be further explored in large population-based datasets, such as those collected by the US Census Bureau, that would also allow for more nuanced comparisons by race-ethnicity.
This study is among the first to explore sex and sexual orientation difference in SES in a nationally representative sample. By using multiple indicators of SES collected by Add Health, our study offers a more comprehensive exploration of SES than has been previously explored in the peer-reviewed literature. Our sexual orientation measure builds on studies that relied on US Census surveys which identified SM on the basis of household composition and focused on same-sex versus different-sex married or cohabitating couples74 75—missing respondents who are single, may not be living with a partner and bisexuals in different-sex relationships. However, as reported above, only 9% and 15.4% of SM women and men, respectively, were living with same-gender partners, suggesting that the SM group identified through a measure that includes a broader array of sexuality options (ie, mostly homosexual, bisexual, mostly heterosexual) identifies a broader group of SM than would be identified through a measure that includes a handful of identity-based options (eg, heterosexual, lesbian or gay, bisexual). These differences in the composition of this SM sample should be considered by readers when comparing findings with studies that used different sexual orientation measures.
Limitations of our study include a reliance on self-report measures; however, we have no reason to suspect systematic reporting bias by sexual orientation. We do not have data on when a SM identity was developed relative to our outcomes and, thus, issues of temporality may impact our results. For instance, models that include respondent education adjust for earlier life differences in SES across groups, which are appropriate if education concluded prior to the development of an SM identity, but may underestimate the effect of SM status on economic status when education was influenced by an individual’s sexual identity. Findings may mask variability in the relationship between sexual orientation and SES across urbanicity and region76; however, exploring these potential differences was beyond the scope of the present study. Findings may also mask variability across gender identity or transgender versus non-transgender (cisgender) status77; however, current gender identity and assigned sex at birth were not collected in Add Health until wave 5 and these new data are not yet available. Lastly, the age of the Add Health cohort (36–44 years) may limit generalisability to other cohorts.
SES is a fundamental contributor to health across the life course26 27 and varies by sexual orientation. Frameworks to analyse sexual orientation inequities in health should consider stigma78 79 and both material and psychosocial pathways to health.33 80 Discrimination, rejection and harassment arise as a consequence of stigma and give rise to what has been termed ‘minority stress’17; however, an over-reliance on Minority Stress theory,1 or on psychosocial theories81 more broadly, to understand population patterns of health will overlook upstream drivers of these conditions. Data gaps should be addressed, specifically, sexual orientation (and gender identity) measures should be added to the Survey of Income and Program Participation, and to administrative systems that track usage of poverty reduction programmes, in order to evaluate the impact of public safety net programmes on the economic status of the population. Future studies should explore the impact of public policies such as marriage, non-discrimination protections and universal health insurance, on earnings and economic status across place and over time, in order to better elucidate the ways in which policies impact SES across sexual orientation groups. Finally, future research should focus on understanding how gender, racial and SM inequality manifests in population patterns of SES at various points in the life course to shed light how and when to intervene to reduce SES inequities and/or to improve the SES of specific population subgroups (eg, SM racial minority women).
What is already known on this subject
Socioeconomic status (SES) is a fundamental contributor to health; however, limited research has examined sexual orientation differences in SES.
Efforts have been hindered by lack of inclusion of sexual orientation identity measures in the nation’s primary sources of economic information about the American public (ie, American Community Survey, Current Population Survey and the Survey of Income and Program Participation), as well as by the limited number of indicators of economic status that are included in population-based health surveys that are beginning to assess sexual orientation.
What this study adds
This study contributes new information about SES by sexual orientation and sex in the nationally representative Add Health sample and provides an agenda for future research on social determinants of observed SES inequities.
Findings indicate that poverty, with accompanying economic strain, is an unappreciated ‘sexual minority’ issue for women.
Among men, lower personal incomes and rates of homeownership, despite higher educational attainment, were observed for sexual minorities.
SES should be considered an important pathway through which sexual orientation health inequities are generated.
Modification analyses suggest that sexual orientation and SES patterns vary between racial minorities (defined as black or Latino) and whites and are different for women and men.
These findings should be replicated in large datasets that allow for more nuanced comparisons across racial-ethnic groups and unpacked.
Future research is needed on upstream SES determinants (eg, marriage, non-discrimination protections, universal healthcare, minimum wage rates, poverty reduction programmes, educational and housing polices) to inform strategies to reduce observed SES inequities along multiple axes of inequality and to improve population health.
The authors would like to thank Ronald R Rindfuss and Barbara Entwisle for assistance in the original design.
Contributors All authors contributed to the conceptualization and writing of the paper. Dr. Conron led the writing and provided overall guidance for the analysis in concert with Dr. Halpern. Dr. Goldberg conducted the analysis.
Funding This study was funded by Eunice Kennedy Shriver National Institute of Child Health and Human Development (10.13039/100009633) and grant number: 3P01HD031921-18S1, 5 R24HD050924, with cooperative funding from 23 other federal agencies and foundations.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.