Abstract
Objectives. To examine differences in HIV prevalence and experiences of discrimination within the trans women community in California’s San Francisco Bay Area.
Methods. Intersectional positions were constructed on the basis of race/ethnicity (non-Hispanic White, non-Hispanic Black, Latina) and gender identity (female identifying, transgender identifying). We used baseline data from the Trans*National study (2016–2017) to construct regression models that estimated racial/ethnic differences in the attribution of discrimination experienced and, along with surrogate measures for intersectionality, estimated risk among those who were dually marginalized (racial/ethnic minority and transgender identifying). Margins plots were used to visually compare absolute risk across all intersectional positions.
Results. Black and Latina trans women were more likely to be HIV positive than non-Hispanic White trans women. In several of the study domains, we estimated a lower risk of reporting discrimination among dually marginalized trans women than among White female-identifying trans women.
Conclusions. Quantitative intersectionality methods highlight the diversity of experiences within the trans women community and reveal potential measurement challenges. Despite facing multiple forms of systemic marginalization, racial/ethnic minority trans women report less discrimination than White trans women. Subjective reporting of discrimination likely undercounts risks among racial/ethnic minorities.
Trans women are a stigmatized group because of their gender identity, and they frequently experience disproportionately poor outcomes with respect to economic security and physical and mental health.1 Trans women are individuals who were assigned male sex at birth and currently identify as women, trans women, or another gender not typically associated with someone who was assigned male sex at birth. Recent studies show a higher likelihood of unmet basic needs (e.g., stable housing) among trans women.2 Trans women are also targets of transphobic discrimination, which has been linked to structural barriers that result in economic insecurity, limited access to health care, psychological distress, and violence.2 Trans women experience a number of health disparities and are disproportionately affected by HIV infection worldwide.3
In San Francisco, California, trans women are burdened with a higher prevalence of HIV infection than any other population.4,5 Trans women in San Francisco have a high mean population viral load, suggesting that, in addition to acquiring HIV at higher rates, they are not receiving optimal medical care.5
Despite the general health and economic vulnerability of this population, trans women are not a monolith, and it is a disservice to examine their health burdens without also considering their diversity. Whereas trans women in San Francisco have a high prevalence of HIV overall, Black women and Latinas have the highest HIV prevalence among trans women,4 respectively accounting for 25% and 27% of the prevalence in this population.6 Moreover, a recent longitudinal study revealed higher rates of HIV acquisition among racial/ethnic minority trans women.7 In addition, although trans women do not identify with the male sex they were assigned at birth, there is wide diversity in their gender identities.6 Gender identity may also be linked to health and economic outcomes. For example, trans women in San Francisco who identify as female are less likely to be HIV positive than those who identify as transgender female.8
Stratified analyses by race/ethnicity or gender identity reveal some disparities but assume homogeneity within these broad categories. Important differences may emerge from the intersection of these demographic categories. Intersectionality provides a framework to examine such differences. Intersectionality examines how interlocking axes of power and privilege on the macro societal level (e.g., racism, sexism, classism) manifest as unique differences in experience for people occupying those intersectional positions on the micro level (e.g., Black trans women, low-income White women).9–12 The experience of people at different intersectional positions is not simply the sum of the risk from their composite identities (i.e., the risk along the racial axis combined with the risk along the gender identity axis); rather, it reflects how these axes of identity and marginalization interact to produce unique outcomes.
This is analogous to statistical interaction, in which the risk among the dually exposed group is greater than (synergistic) or less than (antagonistic) what one would expect from summing the risk from each individual exposure. In contrast with intersectional additivity (in which the risk in the dually exposed group is equivalent to the expectation from simply summing the risk from each individual exposure), this is referred to as intersectional multiplicativity.
Few studies of trans women have examined their health burden through an intersectional lens. Several studies involving intersectional framing have explored how intersectional discrimination (feeling discriminated against because of both one’s racial and gender identities) is associated with experiencing housing discrimination, posttraumatic stress disorder, suicidal ideation, depression, and unequal access to education and employment.2,13 These health and financial outcomes may in turn increase one’s propensity for engaging in sex work, which increases one’s risk for HIV infection.13 These studies have not, however, investigated disparities associated with specific intersectional positions within the trans women community. Highlighting disparities among specific intersectional positions clarifies who are the vulnerable among the vulnerable and informs where targeted strategies are needed to improve health outcomes.
In this study, we applied multiple quantitative intersectionality methods to identify and quantify disparities arising from intersectional positions within the San Francisco trans women community. Analyses incorporating HIV surveillance data indicate that trans women who identify as transgender are at significantly higher risk for HIV than those who identify as female.8 We hypothesized that trans women who are dually marginalized—that is, those who are members of racial/ethnic minority groups and who identify as transgender (rather than female)—would have worse outcomes than those in other intersectional positions (known as the double/multiple jeopardy hypothesis14). In addition, we hypothesized that the prevalence of outcomes among the dually marginalized would demonstrate intersectional multiplicativity. There is currently no consensus on how best to study intersectionality quantitatively.15,16 We therefore employed multiple methods put forward in the literature to showcase their similarities and differences.
METHODS
We used data from the Trans*National Study baseline visit,8 in which HIV prevalence was measured along with sociodemographic correlates of infection. Respondent-driven sampling, a peer-referral sampling strategy, was used to recruit 629 participants in the San Francisco Bay Area.17 Participants were eligible for the study if they were assigned male sex at birth and identified as a gender other than male, were at least 18 years of age, and resided in the San Francisco Bay Area. Data collection for the baseline assessment took place in 2016 and 2017.
Measures
Gender identity was assessed via self-report. Participants were asked “What is your gender identity?” Their response was categorized as one of the following: “male,” “female,” “transgender female or transwoman,” “androgynous/ambigender,” “genderqueer/genderfluid,” “questioning,” or “additional sex or gender.” We restricted our analysis to participants who indicated that their gender was either female (47.2% of the sample) or transgender female or transwoman (49.3% of the sample). Racial/ethnic identity was also assessed via self-report. Participants could indicate more than one race/ethnicity. We restricted this analysis to participants who indicated their race/ethnicity as non-Hispanic White (29% of the sample), non-Hispanic Black/African American (17% of the sample), or Latino/a (32.5% of the sample). The restrictions on gender identity and race/ethnicity yielded an analytic sample of 456 participants.
We used a modified version of the Experiences of Discrimination (EOD) instrument to assess discriminatory experiences in 10 domains (see the Appendix, available as a supplement to the online version of this article at http://www.ajph.org).18,19 Participants were asked “Have you ever experienced discrimination, been prevented from doing something, or been hassled or made to feel inferior because of your gender identity or presentation, or race, ethnicity or color?” Participants were asked similar questions regarding verbal and physical abuse. Each of the EOD domain outcomes were dichotomized. Participants who responded “yes” to experiencing discrimination were then asked whether they believed that discrimination was related to (1) their gender identity or presentation; (2) their race, ethnicity, or color; or (3) both their gender identity and their race/ethnicity. Because it was a primary outcome in the Trans*National Study, we also included laboratory-confirmed HIV status as an outcome in our analysis.
Statistical Analysis
We constructed the following intersectional positions based on self-reported race/ethnicity and gender identity: non-Hispanic White female identifying, non-Hispanic White transgender identifying, non-Hispanic Black female identifying, non-Hispanic Black transgender identifying, Latina female identifying, and Latina transgender identifying. We conducted χ2 analyses to assess differences between the intersectional positions and demographic variables. Multinomial logistic regression models analyzed racial/ethnic differences in the attributions of discrimination reported in the EOD domains. These models included the following 4 outcome levels: (1) no experience of discrimination reported; (2) discrimination attributed to gender identity or presentation; (3) discrimination attributed to race, ethnicity, or color; and (4) discrimination attributed to both gender identity and race/ethnicity.
We used several statistical methods to quantify differences in HIV prevalence and discrimination by intersectional position. Multivariable log-binomial regression models, followed by Stata’s lincom command, estimated the risk difference (RD) between the dually marginalized group (racial/ethnic minority and transgender identifying) and the White female-identifying referent group. We chose White female-identifying participants as the referent group so that our statistical parameters directly quantified risk in the hypothesized dually marginalized intersectional position. If the log-binomial model did not converge, a logistic model was used instead.
We estimated risk differences to calculate interaction on the additive scale. Statistical interaction on the additive scale (not to be confused with intersectional additivity) is consistent with the intersectionality framework because measures on the additive scale directly translate to excess cases of an outcome (that are either caused or prevented) as a result of an exposure or a combination of exposures.16,20 With respect to intersectionality, additive measures translate to excess cases attributed to the synergistic combination of marginalized identities that otherwise would not occur if only one of these exposures were present. Measures on the multiplicative scale do not have this interpretation and can misidentify groups with the greatest health burden because multiplicative measures are dependent on the baseline risk of the outcome in different subgroups.20 All models included age, educational attainment, employment status, and history of incarceration as potential confounders.
In addition, we estimated 4 surrogate measures for intersectionality (Table B, available as a supplement to the online version of this article at http://www.ajph.org). Surrogate measures translate regression parameters on the multiplicative scale to intersectionality quantities.16,21,22 The Synergy Index is the excess risk in the dually marginalized group when there is interaction between exposures relative to the excess risk from either exposure when there is no interaction. The ratio of observed versus expected joint effects on the relative scale (RJE) compares the observed outcome in the dually marginalized group with the counterfactual outcome if there was no interaction effect between exposures. The attributable proportion estimates the proportion of the outcome in the dually marginalized group that is attributed to the intersection or interaction of the exposures. The relative excess risk due to the interaction (RERI) is the excess risk due to the interaction relative to the risk in the referent group (i.e., the risk in each single exposed group).
For each outcome in which either the risk difference or the surrogate measures for the dually marginalized group were statistically significant, we used Stata’s margins function to estimate and plot the predicted probability (or prevalence) of the outcome for each intersectional position. We used Stata 16 in conducting all of our analyses.23 The delta method was used to estimate 95% confidence intervals (CIs) for the surrogate measures.
RESULTS
The mean age of participants was 41 years (range = 18–75 years). The highest educational attainment among a plurality of participants was high school or the equivalent (49.7%). All participants reported annual incomes at or below the poverty line ($28 500; Table 1). The majority of participants (61.4%) reported a history of incarceration (Table 1). There were significant differences across all intersectional positions with respect to education, housing status, history of incarceration, and health insurance (Table 1). Nearly one third of participants were HIV positive, with the highest proportion among Black participants.
TABLE 1—
χ2 Analysis of the Distribution of Sociodemographic Variables by Intersectional Position: Trans*National Study; San Francisco Bay Area, CA; 2016–2017 (n = 456)
| Intersectional Position, No. (%) |
||||||||
| Variable | Full Sample, No. (%) | Non-Hispanic White Female (n = 111) | Non-Hispanic White Transgender Female (n = 54) | Non-Hispanic Black Female (n = 34) | Non-Hispanic Black Transgender Female (n = 70) | Latina Female (n = 78) | Latina Transgender Female (n = 109) | χ2 (P) |
| HIV status | 103.99 (< .005) | |||||||
| Negative | 309 (67.9) | 103 (33.3) | 46 (14.9) | 12 (3.9) | 21 (6.8) | 58 (18.8) | 69 (22.3) | |
| Positive | 146 (32.1) | < 10 (< 6.8) | < 10 (< 6.8) | 22 (15.1) | 49 (33.6) | 20 (13.7) | 39 (26.7) | |
| Age, y | 5.25 (.39) | |||||||
| 18–24 | 46 (10.1) | 10 (21.7) | < 10 (< 21.7) | < 10 (< 21.7) | < 10 (< 21.7) | 10 (21.7) | 11 (23.9) | |
| ≥ 25 | 410 (89.9) | 101 (24.6) | 48 (11.7) | 34 (8.3) | 61 (14.9) | 68 (16.6) | 98 (23.9) | |
| Education | 80.74 (< .005) | |||||||
| High school/equivalent or less | 246 (49.7) | 39 (16.6) | 13 (5.5) | 16 (6.8) | 50 (21.3) | 44 (18.7) | 73 (31.1) | |
| Some college/technical degree | 161 (32.5) | 35 (24.1) | 22 (15.2) | 17 (11.7) | 19 (13.1) | 25 (17.2) | 27 (18.6) | |
| College or higher | 88 (17.8) | 37 (48.7) | 19 (25.0) | < 10 (< 11.4) | < 10 (< 11.4) | < 10 (< 11.4) | < 10 (< 11.4) | |
| Income status | …a | |||||||
| At or below poverty level | 451 (100.0) | 110 (24.4) | 54 (12.0) | 34 (7.5) | 68 (15.1) | 77 (17.1) | 108 (24.0) | |
| Currently employed | 185 (37.7) | 45 (27.1) | 20 (12.1) | 10 (6.0) | 17 (10.2) | 30 (18.1) | 44 (26.5) | 7.23 (.20) |
| Housing status | 28.47 (.002) | |||||||
| Rent/own | 235 (52.0) | 57 (26.6) | 34 (15.9) | 16 (7.5) | 25 (11.7) | 39 (18.2) | 43 (20.1) | |
| Transitional housing | 131 (29.0) | 30 (24.2) | < 10 (< 7.6) | < 10 (< 7.6) | 17 (13.7) | 25 (20.2) | 36 (29.0) | |
| Homeless/shelter | 86 (19.0) | 14 (17.3) | < 10 (< 11.6) | < 10 (< 11.6) | 25 (30.9) | < 10 (< 11.6) | 19 (23.5) | |
| Ever incarcerated | 304 (61.4) | 43 (15.0) | 32 (11.2) | 30 (10.5) | 62 (21.6) | 46 (16.0) | 74 (25.8) | 58.90 (< .005) |
| Type of health insurance | 30.04 (.012) | |||||||
| None | 25 (5.3) | < 10 (< 40.0) | < 10 (< 40.0) | < 10 (< 40.0) | < 10 (< 40.0) | < 10 (< 40.0) | < 10 (< 40.0) | |
| Public | 355 (75.4) | 69 (20.9) | 33 (10.0) | 29 (8.8) | 54 (16.4) | 57 (17.3) | 88 (26.7) | |
| Private | 83 (17.6) | 30 (40.5) | 11 (14.9) | < 10 (< 12.0) | < 10 (< 12.0) | < 10 (< 12.0) | 13 (17.6) | |
| Public and private | 8 (1.7) | < 10 (< 2.2) | < 10 (< 2.2) | < 10 (< 2.2) | < 10 (< 2.2) | < 10 (< 2.2) | < 10 (< 2.2) | |
Note. Female = female identifying; transgender female = transgender identifying. In cells with fewer than 10 observations, true numbers have been masked to protect anonymity.
The χ2 statistic was not applicable because all of the participants were at or below the poverty level.
Results from the multinomial logistic regression analysis indicated that, in nearly all discrimination domains, Black and Latina participants were significantly more likely than White participants to attribute their experiences of discrimination to their intersectional identity (Table 2). White participants were more likely to attribute their experiences of discrimination to their gender identity than either Black or Latina participants (Table 2).
TABLE 2—
Adjusted Multinomial Logistic Regression Analysis of Racial/Ethnic Attribution of Discrimination: Trans*National Study; San Francisco Bay Area, CA; 2016–2017
| Attribution of Discrimination |
||||
| Experience of Discrimination Domain | Gender RRR (95% CI) | Race/Ethnicity RRR (95% CI) | Intersectional Identity RRR (95% CI) | Likelihood Ratio Testa χ2 (P) |
| School | ||||
| Black (Ref = White) | 0.19 (0.09, 0.40) | 4.84 (0.51, 53.34) | 3.89 (1.57, 9.66) | 47.70 (< .005) |
| Latina (Ref = White) | 0.62 (0.37, 1.04) | 5.63 (1.05, 30.29) | 4.74 (2.33, 9.65) | 40.86 (< .005) |
| Latina (Ref = Black) | 3.46 (1.66, 7.21) | 1.20 (0.29, 4.96) | 2.01 (1.17, 3.74) | 13.90 (.003) |
| Getting a job | ||||
| Black (Ref = White) | 0.39 (0.20, 0.77) | 4.39 (0.34, 56.97) | 4.52 (1.58, 12.95) | 24.89 (< .005) |
| Latina (Ref = White) | 0.65 (0.39, 1.07) | 4.39 (0.44, 44.17) | 5.49 (2.47, 12.22) | 35.04 (< .005) |
| Latina (Ref = Black) | 2.09 (1.09, 4.02) | 1.19 (0.22, 6.37) | 1.98 (1.06, 3.72) | 7.22 (.07) |
| At work | ||||
| Black (Ref = White) | 0.48 (0.25, 0.91) | … | 2.20 (0.75, 6.41) | 15.48 (.001) |
| Latina (Ref = White) | 0.71 (0.44, 1.15) | … | 6.08 (2.53, 14.58) | 33.87 (< .005) |
| Latina (Ref = Black) | 2.12 (1.15, 3.92) | 0.34 (0.07, 1.79) | 2.62 (1.31, 5.23) | 14.06 (.003) |
| Getting housing | ||||
| Black (Ref = White) | 0.43 (0.21, 0.89) | … | 2.73 (1.10, 6.74) | 18.99 (< .005) |
| Latina (Ref = White) | 0.87 (0.51, 1.46) | … | 3.63 (1.66, 7.96) | 21.41 (< .005) |
| Latina (Ref = Black) | 2.10 (1.05, 4.20) | 0.71 (0.13, 3.72) | 1.19 (0.63, 2.26) | 5.03 (.17) |
| While staying in a shelter, single-room occupancy, or residential treatment facility | ||||
| Black (Ref = White) | 0.95 (0.49, 1.83) | … | 3.05 (1.08, 8.60) | 6.60 (.09) |
| Latina (Ref = White) | 0.50 (0.28, 0.93) | … | 3.37 (1.41, 8.02) | 20.59 (< .005) |
| Latina (Ref = Black) | 0.47 (0.24, 0.89) | 0.99 (0.08, 11.99) | 1.39 (0.69, 2.79) | 8.52 (.036) |
| Receiving medical care | ||||
| Black (Ref = White) | 0.38 (0.18, 0.81) | … | 1.55 (0.41, 5.84) | 7.59 (.023) |
| Latina (Ref = White) | 0.58 (0.35, 0.96) | … | 4.36 (1.39, 13.69) | 17.92 (< .005) |
| Latina (Ref = Black) | 1.62 (0.78, 3.38) | … | 1.89 (0.76, 4.67) | 7.13 (.07) |
| Getting service in a store or a restaurant | ||||
| Black (Ref = White) | 0.20 (0.09, 0.42) | … | 7.23 (2.57, 20.35) | 60.61 (< .005) |
| Latina (Ref = White) | 0.54 (0.33, 0.87) | … | 5.88 (2.41, 14.37) | 41.95 (< .005) |
| Latina (Ref = Black) | 2.82 (1.37, 5.77) | 0.67 (0.09, 5.18) | 0.82 (0.46, 1.46) | 12.23 (.007) |
| Getting credit, bank loans, or a mortgage | ||||
| Black (Ref = White) | … | … | 7.96 (0.79, 79.88) | … |
| Latina (Ref = White) | 0.20 (0.05, 0.88) | … | 14.97 (1.83, 122.14) | 22.65 (< .005) |
| Latina (Ref = Black) | … | 0.33 (0.04, 2.43) | 1.20 (0.49, 2.90) | 3.95 (.27) |
| On the street or in a public setting | ||||
| Black (Ref = White) | 0.10 (0.04, 0.22) | … | 1.20 (0.52, 2.75) | 60.62 (< .005) |
| Latina (Ref = White) | 0.30 (0.16, 0.56) | … | 1.55 (0.73, 3.31) | 44.16 (< .005) |
| Latina (Ref = Black) | 3.94 (1.97, 7.90) | 2.48 (0.39, 15.92) | 1.84 (1.00, 3.39) | 16.00 (.001) |
| From the police or in court | ||||
| Black (Ref = White) | 0.40 (0.20, 0.79) | … | 6.57 (2.44, 17.67) | 37.94 (< .005) |
| Latina (Ref = White) | 0.64 (0.39, 1.06) | … | 5.54 (2.24, 13.69) | 35.49 (< .005) |
| Latina (Ref = Black) | 1.86 (0.96, 3.59) | 4.85 (0.50, 46.91) | 1.08 (0.59, 1.97) | 5.73 (.13) |
| Experiencing verbal abuse | ||||
| Black (Ref = White) | 0.16 (0.07, 0.38) | … | 0.95 (0.36, 2.48) | 37.458 (< .005) |
| Latina (Ref = White) | 0.28 (0.13, 0.57) | … | 1.75 (0.76, 4.04) | 52.29 (< .005) |
| Latina (Ref = Black) | 2.23 (1.09, 4.57) | 2.77 (0.40, 19.11) | 2.36 (1.16, 4.80) | 6.31 (.10) |
| Experiencing physical abuse | ||||
| Black (Ref = White) | 0.36 (0.19, 0.69) | … | 1.29 (0.57, 2.94) | 13.73 (.001) |
| Latina (Ref = White) | 0.71 (0.43, 1.16) | … | 2.90 (1.46, 5.77) | 25.99 (< .005) |
| Latina (Ref = Black) | 2.10 (1.16, 3.81) | … | 1.94 (1.02, 3.68) | 12.55 (.006) |
Note. CI = confidence interval; RRR = relative risk ratio. The base category was no experience of discrimination. Models adjusted for age, educational attainment, employment status, and history of incarceration. Ellipses indicate that relative risk ratios could not be calculated because there were no White transwomen (the denominator) in corresponding cell.
The likelihood ratio test compared the full model with the model excluding race/ethnicity. A significant P value indicates that the model including the race/ethnicity variable was a better fitting model.
Results were less consistent in comparisons of Black and Latina participants. Relative to Black participants, Latina participants were more likely to attribute discrimination to their gender identity in the domains of school, work, street or public settings, being served at a store or restaurant, and experiencing physical abuse. Latinas were less likely (relative to Black participants) to attribute discrimination experienced while staying in a shelter, single-room occupancy, or residential treatment facility to their gender identity. In addition, Latinas were more likely to attribute discrimination experienced at school, at work, and on the street or in public settings, as well as discrimination when experiencing physical abuse, to their intersectional identity.
Multivariable regression analyses indicated several outcomes in which dually marginalized intersectional positions were significantly different from the White female-identifying reference group. Non-Hispanic Black transgender-identifying participants had a 52% increased risk of testing HIV positive (RD = 0.52; 95% CI = 0.37, 0.67) relative to White female-identifying participants (Table 3). However, in comparison with the reference group, non-Hispanic Black transgender-identifying participants had an 81% decreased risk of reporting feeling discriminated against when receiving medical care (RD = −0.19; 95% CI = −0.32, −0.06), an 82% decreased risk of reporting verbal abuse (RD = −0.18; 95% CI = −0.32, −0.04), and an 83% decreased risk of reporting physical abuse (RD = −0.17; 95% CI = −0.34, −0.01; Table 3). Relative to White female-identifying participants, Latina transgender-identifying participants had a 24% increased risk of testing HIV positive (RD = 0.24; 95% CI = 0.14, 0.34) (Table 3).
TABLE 3—
Intersectionality Analysis Quantifying Risk Among Dually Marginalized Participants: Trans*National Study; San Francisco Bay Area, CA; 2016–2017
| Surrogate Measure |
||||
| Outcome | Race × Gender RD | RJE | AP | RERI |
| Black transgender identifying | ||||
| HIV status | 0.52 (0.37, 0.67)a | 1.20 (0.03, 2.38) | 0.17 (−0.65, 0.98) | 3.54 (−14.90, 21.99) |
| EOD | ||||
| At school | −0.00 (−0.17, 0.17) | 0.77 (−0.00, 1.53) | −0.31 (−1.61, 1.00) | −0.30 (−1.56, 0.96) |
| Getting a job | −0.02 (−0.19, 0.15) | 0.92 (0.41, 1.42) | −0.09 (−0.70, 0.51) | −0.09 (−0.67, 0.49) |
| At work | −0.02 (−0.19, 0.15) | 22.22 (−333.32, 3777.75) | 0.95 (0.23, 1.68) | 0.86 (0.19, 1.53) |
| Getting housing | 0.04 (−0.12, 0.21) | 0.82 (0.03, 1.60) | −0.22 (−1.39, 0.95) | −0.26 (−1.59, 1.08) |
| While staying in a shelter, single-room occupancy, or residential treatment facility | 0.01 (−0.17, 0.19) | 0.96 (0.46, 1.45) | −0.05 (−0.59, 0.50) | −0.05 (−0.68, 0.58) |
| Receiving medical care | −0.19 (−0.32, −0.06) | 0.39 (0.04, 0.74) | −1.57 (−3.86, 0.73) | −0.60 (−1.34, 0.13) |
| Store/restaurant | 0.04 (−0.13, 0.20) | 1.71 (0.48, 2.94) | 0.42 (−0.00, 0.83) | 0.42 (−0.02, 0.87) |
| Getting credit, bank loans, mortgage | 0.04 (−0.12, 0.20)a | 0.55 (−0.04, 1.14) | −0.82 (−2.76, 1.12) | −1.04 (−3.35, 1.26) |
| Street/public settings | −0.14 (−0.30, 0.01)a | 0.35 (−0.19, 0.88) | −1.88 (−6.34, 2.57) | −0.85 (−2.97, 1.28) |
| Police/court | −0.01 (−0.18, 0.16) | 0.90 (0.50, 1.31) | −0.11 (−0.61, 0.39) | −0.10 (−0.59, 0.38) |
| Verbally abused | −0.18 (−0.32, −0.04)a | 0.11 (−0.08, 0.29) | −8.19 (−23.82, 7.45) | −2.40 (−7.55, 2.76) |
| Physically abused | −0.17 (−0.34, −0.01) | 0.72 (0.38, 1.05) | −0.39 (−1.04, 0.26) | −0.28 (−0.73, 0.17) |
| Latina transgender identifying | ||||
| HIV status | 0.24 (0.14, 0.34)a | 1.23 (0.33, 2.14) | 0.19 (−0.40, 0.78) | 1.11 (−2.60, 4.82) |
| EOD | ||||
| At school | 0.13 (−0.01, 0.27) | 0.78 (0.22, 1.33) | −0.29 (−1.21, 0.63) | −0.49 (−1.99, 1.02) |
| Getting a job | 0.08 (−0.06, 0.22) | 1.18 (0.70, 1.65) | 0.15 (−0.19, 0.49) | 0.17 (−0.22, 0.57) |
| At work | 0.05 (−0.08, 0.19) | 2.21 (−0.53, 4.95) | 0.55 (−0.01, 1.11) | 0.73 (−0.08, 1.53) |
| Getting housing | 0.12 (−0.03, 0.26) | 0.99 (0.53, 1.47) | −0.00 (−0.48, 0.47) | −0.00 (−0.62, 0.61) |
| While staying in a shelter, single-room occupancy, or residential treatment facility | −0.07 (−0.22, 0.09) | 0.90 (0.39, 1.41) | −0.11 (−0.73, 0.52) | −0.09 (−0.64, 0.45) |
| Getting medical care | −0.05 (−0.18, 0.09) | 0.89 (0.33, 1.44) | −0.13 (−0.84, 0.58) | −0.11 (−0.70, 0.49) |
| Store/restaurant | 0.06 (−0.08, 0.20) | 1.50 (0.73, 2.28) | 0.33 (−0.01, 0.68) | 0.36 (−0.01, 0.73) |
| Getting credit, bank loans, mortgage | 0.10 (−0.03, 0.23)a | 0.95 (0.03, 1.87) | −0.05 (−1.08, 0.97) | −0.10 (−2.02, 1.81) |
| Street/public settings | −0.05 (−0.17, 0.06)a | 0.48 (−0.18, 1.14) | −1.08 (−3.93, 1.77) | −0.76 (−2.79, 1.28) |
| Police/court | 0.02 (−0.12, 0.16) | 1.00 (0.62, 1.40) | 0.01 (−0.38, 0.40) | 0.01 (−0.41, 0.43) |
| Verbally abused | −0.09 (−0.19, 0.00) | 0.17 (−0.12, 0.46) | −4.97 (−15.29, 5.34) | −2.10 (−6.65, 2.45) |
| Physically abused | 0.02 (−0.12, 0.15) | 0.96 (0.64, 1.29) | −0.04 (−0.39, 0.32) | −0.04 (−0.40, 0.32) |
Note. AP = attributable proportion; EOD = experience of discrimination; RD = risk difference; RERI = relative excess risk due to interaction; RJE = ratio of observed to expected joint effects. Models controlled for age, educational attainment, employment status, and history of incarceration. The null value for the RJE is 1, and the null value for the AP and RERI is 0. The reference for group is non-Hispanic White female identifying.
The log-binomial model did not converge, so logistic models (odds ratios) were used instead.
Surrogate measures provided evidence for intersectional multiplicativity in several domains (Table 3). Black transgender-identifying participants exhibited synergistic interaction when reporting discrimination at work (AP = 0.95; 95% CI = 0.23, 1.68; RERI = 0.86; 95% CI = 0.19, 1.53). We observed antagonistic interaction among Black transgender-identifying participants for reports of discrimination when receiving medical care (RJE = 0.39; 95% CI = 0.04, 0.74), discrimination on the street and in public settings (RJE = 0.35; 95% CI = −0.19, 0.88), and reports of verbal abuse (RJE = 0.11; 95% CI = −0.08, 0.29). Among Latina transgender-identifying participants, we observed antagonistic interaction for reports of verbal abuse (RJE = 0.17; 95% CI = −0.12, 0.46). Synergy Index estimates were inconsistent with the remaining 3 surrogate measures (suggesting interaction in the opposite direction) because one or both “exposures” were statistically preventative in bivariate analyses.20 Therefore, these estimates are not reported here.
Figure 1 provides a visualization (for each intersectional position) of the absolute risk of testing HIV positive or reporting discriminatory experiences in 6 domains. These margins plots indicate that the risk of reporting any verbal abuse or physical abuse is largely ubiquitous across intersectional positions. The plots also highlight that Black participants overall (not only those who are dually marginalized) are at elevated risk for testing HIV positive.
FIGURE 1—
Predicted Probabilities of HIV-Positive Status or Reports of Discrimination, by Intersectional Position, in the Domains (a) HIV Positive, (b) at Work, (c) Receiving Medical Care, (d) at a Store or Restaurant, (e) on the Street or in Public, and (f) Experiencing Physical Abuse: Trans*National Study; San Francisco Bay Area, CA; 2016–2017
Note. F = female identifying; TG = transgender identifying.
DISCUSSION
We investigated disparities in health and social outcomes among trans women in San Francisco through the lens of intersectionality. Our findings showed that there is variability in the perception of discrimination throughout the trans women community. Trans women of color were more likely than White trans women to perceive dual marginalization, attributing their discriminatory experiences to both their gender and racial/ethnic identities. A comparison of Latina trans women and Black trans women showed that Latinas were often more likely to attribute discriminatory experiences to their gender identity and their intersectional identity.
These results necessitate further investigation into how systematic marginalization operates in different domains and is differentially experienced among trans women of color. Cumulative experiences of intersectional discrimination are associated with an increased risk of housing instability.2 Considering the variability in attributions of discrimination, it is worth investigating whether (among those reporting discrimination) attribution of discrimination also moderates the association with housing instability and other health outcomes.
Our results did not support the double jeopardy hypothesis. Dually marginalized trans women did not have a greater risk of reporting discrimination than White female-identifying trans women. Perhaps counterintuitively, dually marginalized trans women were significantly less likely to report discrimination than White female-identifying trans women in the domains of receiving medical care, experiencing verbal abuse, and experiencing physical abuse. Although both Black and Latina transgender-identifying participants were more likely to test HIV positive than White female-identifying participants, the margins plot reveals that HIV-positive status is correlated with race and is not unique to this intersectional position. In addition, surrogate measures did not provide evidence that this increased risk of HIV-positive status among those who were dually marginalized was attributable to the interacting effects of race/ethnicity and gender identity.
There were several instances in which we found support for intersectional multiplicativity with surrogate measures. Relative to the counterfactual in which there is no interacting effect of race/ethnicity and gender identity, non-Hispanic Black transgender-identifying participants were 61% less likely to report experiencing discrimination when receiving medical care, 65% less likely to report being discriminated against on the street or in public settings, and 89% less likely to report ever having been verbally abused. Similarly, Latina transgender-identifying participants were 83% less likely to report being verbally abused. In one domain, discrimination at work, non-Hispanic Black transgender-identifying participants were more likely to report discrimination relative to the counterfactual.
Collectively, our results are surprising in light of the multiple forms of systemic marginalization and disproportionate violence faced by racial/ethnic minorities24 and, specifically, trans women of color.6,25–28 Notably, the discrimination questions were a measure of participants reporting discrimination they had experienced; these questions were not an objective measurement of discrimination experienced.
In the discrimination research literature, it is not uncommon for White participants to report more discrimination than racial/ethnic minority participants.29,30 In a recent analysis of trans women in San Francisco involving data from 2013, Arayasirikul et al. found that White trans women reported greater transphobic discrimination than trans women of color.31 Also, using data from a nationally representative study of adults receiving HIV care, Baugher et al. found a higher prevalence of self-reported discrimination in health care settings among White participants than among Latino and Black participants.32 As with our analysis, neither of these studies claimed that White participants experience more discrimination than racial/ethnic minority participants; rather, White participants are more likely to report experiencing discrimination.
“Ceiling effects” are one potential explanation proposed in the intersectionality literature for why non-Hispanic White participants may report more negative experiences (with added marginalized social identities) than members of racial/ethnic minority groups.33 According to this explanation, racial/ethnic minorities endure a higher baseline level of victimization such that the perception of further victimization from additional marginalized identities is minimal. Relatedly, other scholars noting this phenomenon have hypothesized that discrimination may be so prevalent among certain groups that it is expected and therefore underreported as noteworthy24,29 or that it may be acknowledged on the group level but minimized on the personal level.34 Our results may highlight the need for a more objective measure of discrimination that is sensitive enough to register experiences of discrimination among populations with (potentially) different reference points.
Limitations
Our results should be interpreted in the context of several limitations. First, the relatively small sample sizes for each intersectional position resulted in wide confidence intervals, and there may not have been sufficient power to observe statistically significant differences. Sample size considerations also prevented us from analyzing interactions with other racial minority groups such as Asians and Native Americans. Second, individuals could have been misclassified by race/ethnicity. Anyone who indicated Latino/a ethnicity was coded as “Latina,” including Black Latinas. Our results remained largely unchanged when we reclassified our 25 Black Latinas as Black; however, several estimates shifted in statistical significance (Table A, available as a supplement to the online version of this article at http://www.ajph.org).
Third, discrimination outcomes were self-reported and subjective. There may be differences between racial/ethnic groups regarding the level of victimization that meets the threshold of being considered discriminatory. Fourth, we restricted our intersectional analysis to race/ethnicity and gender identity because these were the identities that were the subject of the discrimination questions. Other social identities and characteristics relating to societal power could have been included as well to construct intersectional positions (e.g., language, immigration status, housing status, and “passing” as cisgender female). However, constructing additional intersectional positions would have magnified our concerns about sample size and statistical power.
Finally, all of the participants reported an annual income at or below the poverty line. Alongside other sociodemographic indicators, this may signal that socioeconomic disadvantage was prevalent in our sample, potentially obscuring differences according to intersectional position. The prevalence of socioeconomic disadvantage may also indicate that our sample is not generalizable to the broader community of trans women in San Francisco. Although respondent-driven sampling is a common probability-based sampling strategy used with marginalized populations (such as trans women), the sampling process tends to reach the more socioeconomically disadvantaged segments of the population (relative to alternative approaches such as time location sampling).35 This potential sampling bias likely limits the generalizability of our findings.
Conclusions
Our analysis provides insight into the diversity of experiences within the transgender community. We have demonstrated the use of multiple quantitative methods, individually and in combination, to study potential differences by intersectional position. Calculating the risk difference for the dually marginalized group with respect to White female-identifying participants allowed us, in part, to test the double jeopardy hypothesis but limited us to making comparisons with a single referent category. Estimating surrogate measures quantifies how much of the risk in the dually marginalized group is beyond what we would expect from the counterfactual scenario in which there are no interaction effects. These measures provide an indication of intersectional multiplicativity but fall short of contextualizing how risk in the dually marginalized group compares with risk in other intersectional positions.
Margins plots provide an assessment of absolute risk across all intersectional positions, with statistically significant differences noted via error bars. These plots are easily interpretable and illustrate where disparities exist without making any group the reference or centering it as the standard for comparison. Such an approach, as Bauer noted of intersectionality broadly, is useful for providing a “precise identification of inequalities.”16(p11)
Experiences of discrimination are prevalent among trans women in San Francisco. Racism and transphobic discrimination have been linked to risk factors for HIV infection on both the individual level and the structural level. Intervening on systemic discrimination is a potential population-based course of action to interrupt HIV transmission and offset other adverse health outcomes for trans women. We used quantitative intersectionality methods to identify high-risk outcomes ubiquitous to the entire population (e.g., verbal abuse), higher-risk outcomes along a single axis of identity (e.g., HIV prevalence among Black participants), and outcomes specific to an intersectional position (e.g., reduced risk of reporting discrimination when receiving medical care among Black transgender-identifying participants). With sufficient sample sizes and objective measures, these methods can be useful in determining when interventions should be applied to a broad population or when they should be targeted to a specific intersectional group.
ACKNOWLEDGMENTS
The Trans*National Study was made possible through funding from the National Institute of Minority Health and Health Disparities (grant R01 MD010678/A133681). P. W. was also supported by funding from the National Institute of Allergy and Infectious Diseases (grant K01 AI145572).
We are grateful to our study participants for giving us their time and for sharing with us their lived experiences.
CONFLICTS OF INTEREST
The authors have no conflicts of interest to declare.
HUMAN PARTICIPANT PROTECTION
The Trans*National Study was approved by the institutional review board of the University of California, San Francisco.
Footnotes
See also Biello and Hughto, p. 344.
REFERENCES
- 1.Winter S, Diamond M, Green J et al. Transgender people: health at the margins of society. Lancet. 2016;388(10042):390–400. doi: 10.1016/S0140-6736(16)00683-8. [DOI] [PubMed] [Google Scholar]
- 2.Beltran T, Allen AM, Lin J, Turner C, Ozer EJ, Wilson EC. Intersectional discrimination is associated with housing instability among trans women living in the San Francisco Bay Area. Int J Environ Res Public Health. 2019;16(22):4521. doi: 10.3390/ijerph16224521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Poteat T, Scheim A, Xavier J, Reisner S, Baral S. Global epidemiology of HIV infection and related syndemics affecting transgender people. J Acquir Immune Defic Syndr. 2016;72(suppl 3):210–219. doi: 10.1097/QAI.0000000000001087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rapues J, Wilson EC, Packer T, Colfax GN, Raymond HF. Correlates of HIV infection among transfemales, San Francisco, 2010: results from a respondent-driven sampling study. Am J Public Health. 2013;103(8):1485–1492. doi: 10.2105/AJPH.2012.301109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Santos G, Wilson EC, Rapues J, Macias O, Packer T, Raymond HF. HIV treatment cascade among transgender women in a San Francisco respondent driven sampling study. Sex Transm Infect. 2014;90(5):430–433. doi: 10.1136/sextrans-2013-051342. [DOI] [PubMed] [Google Scholar]
- 6.Wilson EC, Chen Y, Arayasirikul S et al. Differential HIV risk for racial/ethnic minority trans*female youths and socioeconomic disparities in housing, residential stability, and education. Am J Public Health. 2015;105(suppl 3):e41–e47. doi: 10.2105/AJPH.2014.302443. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.McFarland W, Wesson PD, Turner C et al. High HIV incidence among young and racial/ethnic minority transgender women in San Francisco: results of a longitudinal cohort study. J Acquir Immune Defic Syndr. 2020;84(1):e7–e10. doi: 10.1097/QAI.0000000000002315. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Raymond HF, Wilson EC, Packer T, Ick T, Lin J, Mcfarland W. High and stable human immunodeficiency virus prevalence among transwomen with low income recruited with respondent-driven sampling, San Francisco, 2010–2016. Sex Transm Dis. 2019;46(2):118–124. doi: 10.1097/OLQ.0000000000000916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bowleg L. The problem with the phrase women and minorities: intersectionality—an important theoretical framework for public health. Am J Public Health. 2012;102(7):1267–1273. doi: 10.2105/AJPH.2012.300750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cho S, Chrenshaw KW, McCall L. Toward a field of intersectionality studies: theory, applications, and praxis. Signs J Women Cult Soc. 2013;38(4):785–810. doi: 10.1086/669608. [DOI] [Google Scholar]
- 11.Crenshaw K. Mapping the margins: intersectionality, identity politics, and violence against women of color. Stanford Law Rev. 1991;43(6):1241–1299. doi: 10.2307/1229039. [DOI] [Google Scholar]
- 12.Crenshaw K. Demarginalizing the intersection of race and sex: a black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics. Univ Chic Leg Forum. 1989;1:8. [Google Scholar]
- 13.Wilson EC, Sean YC, Fisher AH, Mcfarland W. The impact of discrimination on the mental health of trans*female youth and the protective effect of parental support. AIDS Behav. 2016;20(10):2203–2211. doi: 10.1007/s10461-016-1409-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Vu M, Li J, Haardörfer R, Windle M, Berg CJ. Mental health and substance use among women and men at the intersections of identities and experiences of discrimination: insights from the intersectionality framework. BMC Public Health. 2019;19(108):1–13. doi: 10.1186/s12889-019-6430-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Nash JC. Re-thinking intersectionality. Fem Rev. 2008;89(1):1–15. doi: 10.1057/fr.2008.4. [DOI] [Google Scholar]
- 16.Bauer GR. Incorporating intersectionality theory into population health research methodology: challenges and the potential to advance health equity. Soc Sci Med. 2014;110:10–17. doi: 10.1016/j.socscimed.2014.03.022. [DOI] [PubMed] [Google Scholar]
- 17.Heckathorn D. Respondent-driven sampling: a new approach to the study of hidden populations. Soc Probl. 1997;44:174–199. [Google Scholar]
- 18.Krieger N. Racial and gender discrimination: risk factors for high blood pressure? Soc Sci Med. 1990;30(12):1273–1281. doi: 10.1016/0277-9536(90)90307-E. [DOI] [PubMed] [Google Scholar]
- 19.Krieger N, Smith K, Naishadham D, Hartman C, Barbeau EM. Experiences of discrimination: validity and reliability of a self-report measure for population health research on racism and health. Soc Sci Med. 2005;61(7):1576–1596. doi: 10.1016/j.socscimed.2005.03.006. [DOI] [PubMed] [Google Scholar]
- 20.VanderWeele TJ, Knol MJ. A tutorial on interaction. Epidemiol Methods. 2014;3(1):33–72. doi: 10.1515/em-2013-0005. [DOI] [Google Scholar]
- 21.Skrondal A. Interaction as departure from additivity in case-control studies: a cautionary note. Am J Epidemiol. 2003;158(3):251–258. doi: 10.1093/aje/kwg113. [DOI] [PubMed] [Google Scholar]
- 22.Jackson JW, Williams DR, Vanderweele TJ. Disparities at the intersection of marginalized groups. Soc Psychiatry Psychiatr Epidemiol. 2016;51(10):1349–1359. doi: 10.1007/s00127-016-1276-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC; 2019. [Google Scholar]
- 24.van Ryn M, Fu SS. Paved with good intentions: do public health and human service providers contribute to racial/ethnic disparities in health? Am J Public Health. 2003;93(2):248–255. doi: 10.2105/AJPH.93.2.248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Wilson EC, Arayasirikul S, Johnson K. Access to HIV care and support services for African American transwomen living with HIV. Int J Transgend. 2013;14(4):182–195. doi: 10.1080/15532739.2014.890090. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mealer B. “It’s like we’re not even human”: the reality of being a trans black woman in America. Available at: https://www.theguardian.com/lifeandstyle/2019/jun/27/texas-trans-black-women-murders-reality. Accessed December 19, 2020.
- 27.Stafford Z. Transgender homicide rate hits historic high in US, says new report. Available at: https://www.theguardian.com/us-news/2015/nov/13/transgender-homicide-victims-us-has-hit-historic-high. Accessed December 19, 2020.
- 28.James SE, Herman JL, Rankin S, Keisling M, Mottet L, Anafi M. The report of the 2015 US Transgender Survey. Available at: https://transequality.org/issues/us-trans-survey. Accessed December 19, 2020.
- 29.Schuster MA, Collins R, Cunningham WE et al. Perceived discrimination in clinical care in a nationally representative sample of HIV-infected adults receiving health care. J Gen Intern Med. 2005;20(9):807–813. doi: 10.1111/j.1525-1497.2005.05049.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Krieger N, Sidney S. Prevalence and health implications of anti-gay discrimination: a study of black and white women and men in the CARDIA cohort. Int J Health Serv. 1997;27(1):157–176. doi: 10.2190/HPB8-5M2N-VK6X-0FWN. [DOI] [PubMed] [Google Scholar]
- 31.Arayasirikul S, Wilson EC, Raymond HF. Examining the effects of transphobic discrimination and race on HIV risk among transwomen in San Francisco. AIDS Behav. 2017;21(9):2628–2633. doi: 10.1007/s10461-017-1728-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Baugher AR, Beer L, Fagan JL et al. Discrimination in healthcare settings among adults with recent HIV diagnoses. AIDS Care. 2019;31(9):1077–1082. doi: 10.1080/09540121.2018.1545988. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cooper Stoll L, Block R. Intersectionality and cyberbullying: a study of cybervictimization in a midwestern high school. Comput Human Behav. 2015;52:387–397. [Google Scholar]
- 34.Taylor DM, Wright SC, Moghaddam FM, Lalonde RN. The personal/group discrimination discrepancy: perceiving my group, but not myself, to be a target for discrimination. Pers Soc Psychol Bull. 1990;16(2):254–262. doi: 10.1177/0146167290162006. [DOI] [Google Scholar]
- 35.Wei C, Mcfarland W, Colfax GN, Fuqua V, Raymond HF. Reaching black men who have sex with men: a comparison between respondent-driven sampling and time-location sampling. Sex Transm Infect. 2012;88(8):622–626. doi: 10.1136/sextrans-2012-050619. [DOI] [PMC free article] [PubMed] [Google Scholar]