5
Measuring Sexual Orientation and Identifying Sexual Minority Populations
This chapter reviews approaches to assessing sexual orientation in large-scale population-based settings. We begin with an overview of the construct of sexual orientation and an explanation of how the measures we highlight in this report relate to the broader construct. Following this brief overview, we review existing approaches for assessing sexual orientation identity as an indicator of sexual orientation, allowing for the enumeration of sexual majority and sexual minority populations. We then offer our recommendations for measurement practices to document sexual orientation identities, and we conclude with recommendations for future research.
As noted in Part I, sexual orientation is a multidimensional construct reflecting in some way the gender(s) of a person’s desired potential or actual sexual or romantic partners relative to the person’s own gender.1 The measurement of sexual orientation has varied across settings and data collection purposes, with a focus on one or a combination of the three dimensions of sexual orientation: behavior, attraction, and identity (van Anders, 2015; Tolman at al., 2014; Diamond, 2003; Coleman, 1987). Given the complexities of categorizing sexual behaviors and erotic or romantic attractions across a range of diverse gender identities as indicators of sexual orientation (van Anders, 2015), there are unique challenges to assessing sexual orientation through the dimensions of behavior and attraction.
___________________
1 Throughout this chapter we use “gender” rather than “gender identity” because, while sexual orientation may be based on gender identity, it may also consider other dimensions of gender, such as gender expression.
SEXUAL ORIENTATION DIMENSIONS: CONCEPTS AND MEASUREMENT
The construct of sexual behavior is often characterized as the sexually expressed component through physical and nonphysical expression. Sexual behavior can also include sexual abstinence, celibacy, or not yet physically expressed behaviors independent of sexual orientation identity. For example, one can self-identify as gay but decide to be celibate, or one can self-identify lesbian but not yet have experienced sexual physical contact with a woman. Primarily, collecting data on sexual behavior has served to inform public health epidemiological surveillance and intervention, health care practice, and ethnographic assessment of sexuality (Brooks et al., 2018; Wolff et al., 2017; Sell, 1997).
Identifying all of the possible sexual behavior measures and types is beyond the scope of this report. Often, measures of sexual behavior (e.g., number of partners, occasions of engaging in a given sexual act) are selected to meet a specific purpose, such as estimating the risks of pregnancy, HIV, or sexually transmitted infections or understanding or distinguishing among diverse sexual partnerships. Questionnaire design typically involves elaborate skip and filtering instructions (see Fenton et al., 2001) particularly related to partner information—including multiple sexual partners—as well as anatomical information of both respondents and partners (see Webb et al., 2015; Ivankovich, Leichliter, and Douglas, 2013; Fenton et al., 2001).
Similarly, approaches to the measurement of sexual attraction as an indicator of sexual orientation center on categorizing people by the gender(s) to which they are attracted, sexually or romantically. The measurement of sexual attraction may be appropriate when the purpose of the data collection is to assess the complex and fluid ways that individuals may experience attraction across a range of genders and relationships, irrespective of whether they enact these attractions into behavior or how they self-identify (Wolff et al., 2017; Diamond, 2003). Although the study of erotic and romantic attractions can serve many important data collection goals, in administrative setting and general population surveys, measures of sexual attraction have often been included as an indicator of sexual orientation when there are concerns that stigma and prejudice may inhibit individuals’ self-disclosure of their sexual orientation identity.
The measurement of sexual attraction has also been used to characterize individuals who are not or not yet sexually active, as done in the National Longitudinal Study of Adolescent to Adult Health (Add Health) (Harris et al., 2009). Approaches include providing response options across a range of genders (men, women, men and women, etc.) or a range of sexual orientation terms that imply the gender to which someone is attracted (e.g., mostly homosexual). Over the years, the measurement of sexual attraction
has varied, alongside the terms used to define interrelated phenomena, such as sexual arousal, sexual feelings, sexual desire, sexual attraction, and erotic attraction (for reviews, see Patterson et al., 2017; Sexual Minority Assessment Research Team, 2009; Diamond, 2003; Sell, 1997).
The final dimension of sexual orientation, sexual orientation identity, is often characterized as reflecting a person’s sense of self with regard to enduring romantic and sexual attraction to particular gender(s) (Sell, 1997; Laumann et al., 1994). Sexual orientation identity is the cognitive as well as social expression of one’s sexual orientation. Thus, sexual orientation identity is the dimension that is most consistently tied to experiences with material forms of discrimination (Drydakis, 2022; Pachankis et al., 2020; Dillbary and Edwards, 2019; Pachankis and Bränström, 2018) and noted explicitly in laws and policies aimed at protecting or harming sexual minorities (e.g., U.S. Equal Employment Opportunity Commission, 2021; U.S. Department of Justice, 2015). It is also the dimension with the broadest and longest use in population-based data collection settings (surveys and administrative data) to enumerate and distinguish among sexual minority and majority populations (Patterson et al., 2017; Federal Interagency Working Group, 2016b). Given the implications for measuring identity when tracking services and outcomes for sexual minority populations and the long-standing inclusion of measures of sexual orientation identity as a tool for assessing sexual orientation, the remainder of the chapter focuses on a review and recommendation of measurement approaches of sexual orientation identity.
EVALUATION OF EXISTING MEASUREMENT APPROACHES
The panel considered existing sexual orientation identity measures included in federal and national surveys and reviewed the prior recommendations published in peer-reviewed journals (e.g., Patterson et al., 2017), in books (e.g., Stall et al., 2020), and in policy briefs and reports (e.g., Federal Interagency Working Group, 2016a; The Gender Identity in U.S. Surveillance Group [GenIUSS], 2014; SMART, 2009) and well-respected think tanks and academic research centers (e.g., the UCLA Williams Institute, the National Opinion Research Center). Tables 5A-1 and 5A-2, in the annex to this chapter, list measures that the panel evaluated based on the measurement criteria outlined in Chapter 4 and our principles for data collection: inclusiveness, precision, autonomy, parsimony, and privacy (see Chapter 2).
Measures of sexual orientation identity have not varied dramatically over the last three decades. There are three general types of approaches: (1) asking whether someone identifies as a sexual minority or LGBT with dichotomous “yes/no” response options; (2) asking a respondent to select from a set of options reflecting combinations of sexual orientation identity
and attraction terminologies (e.g., mostly homosexual [gay], but somewhat attracted to opposite sex); and (3) asking a respondent to select from a set of sexual orientation identity labels or terms (e.g., lesbian, straight, bisexual). The dichotomous response approach used by the Gallup Poll provided one of the first and strongest population-based estimates of the total LGBT population in the United States. However, it did not allow for the distinct measurement of sexual orientation separately from gender identity. The response option set, which includes a mix of terminology reflecting attraction and identity, appear to be more common among population-based data collection targeting younger respondents, such as Add Health and the National Survey of Youth in Custody. Most of the measures assessing sexual orientation identity use the third approach in which the focus is documenting which sexual orientation identity labels people use to describe themselves, with slight variations in stems and response options.
Across the surveys assessing identification with sexual orientation labels, slightly different wording is used in the question stems, such as “do you consider yourself,” “describe your sexual orientation,” or “think of yourself.” The panel weighed concerns related to our principle of precision over whether naming the construct “sexual orientation” or “sexual identity” in the question stem was required for adequate performance of the measure. Avoiding the use of technical and theoretical language is a best practice in measurement design, and many scholars have considered the use of the term “sexual orientation” in general population data collection to be too technical (Badgett et al., 2009). Recent qualitative research has suggested that some respondents would prefer the specification of the concept they are being asked about, and yet the study also found that participants ranged in their understanding of “sexual orientation” (Suen et al., 2020). Moreover, in this study, the participants were mostly highly educated, and those who reported a desire to have the term “sexual orientation” specified in the stem were White and identified as a sexual minority, a demographic that represents a narrow subset of the general U.S. population. As such, it appears that while it may seem more precise to add the construct terminology to the stem of the question about sexual orientation identity, it is not yet clear whether it would produce more precise and accurate estimates of sexual orientation.
Additionally, the response options differ slightly in terms of whether a “none of the above” type of option is listed, whether a free text option is provided, and the range of identities offered for selection. Terms used to affirmatively describe people who identify with a sexual orientation indicating a sexual minority status have shifted throughout U.S. history, although some terms have remained popular for decades. “Gay,” “lesbian,” and “bisexual” have been present in literature and community organizing among sexual minorities since the early to mid-1900s (Morris, 2009). Other
terms have been asserted in response or resistance to what is considered the mainstream Eurocentric sexual minority community, including Two-Spirit, same gender loving, and queer. There are also varying terms for sexual orientation identity that reflect plurisexual attractions and behaviors (e.g., pansexual, queer, omnisexual) and terms used to reflect the lack of erotic attraction (e.g., grey, asexual) (Jourian, 2015; Lassiter, 2014; Galupo et al., 2014).
The placement of sexual orientation identity terms varies across data collection contexts and has changed over time. In several early versions of surveys including measures of sexual orientation identity, the question was placed near questions about sexual risk and health care (e.g., in the National Health and Nutrition Examination Survey [NHANES], and the National Survey of Family Growth [NSFG]) or in sections about general life issues (the National Health Interview Survey [NHIS]) (Dahlhamer et al., 2014). The current 2021 NHIS also includes a sexual orientation identity item in the context of questions about social support and relationship status. However, other surveys have had success including it with the demographic items without increasing nonresponse. For example, the Census Pulse Survey, which introduced a question on sexual orientation identity in 2021, includes the item as the eighth question overall following the two-step question on transgender identity (see Chapter 6) and before asking about household size and marital status. However, the placement of sexual orientation identity questions in a survey merits further assessment of its effect on response rates and the response options selected.
In reviewing the available data for each measure’s performance, the panel also considered several other factors: the empirical evidence (e.g., cognitive interviews, psychometric evaluation, nonresponse rates) supporting its use in large-scale population studies; its use in both probabilistic and nonprobabilistic sampling frames; the extent to which it has been used in diverse populations (e.g., across races, ages, sexual and gender minority populations, regions); and the extent to which it has been used with different modes of data collection (e.g., interviewer administered, paper-and-pencil administration, computer-assisted interviews with or without audio, web based). The panel also considered whether the measure was indicative of sexual orientation identity and did not conflate its use with other dimensions of sexual orientation (i.e., indicators of attraction or sexual behavior).
Evidence of Measure Performance in the United States
The panel explored whether the existing sexual orientation identity measures use a stem and response options with strong face construct validity (see above and Chapters 1 and 4). We prioritized the recommendation of measures of sexual orientation identity that reflect both the cognitive
dimension of identity (i.e., how a respondent sees or thinks of themselves) and the social or political dimension of identity (i.e., provides label options that reflect a social status or community). We also assessed whether the measure met the content standard by including a stem and response options that reflected a focus on sexual orientation identity labels separate from gender identity, sexual behavior, or attraction. Only the measures asking respondents to select an identity label, which we refer to as the “identity label approach,” fit this criterion and therefore became the focus of the rest of our evaluation and discussion of a recommended measure of sexual orientation identity.
The sexual orientation identity label approach has undergone a significant amount of cognitive interview and survey design testing to assess whether respondents understood the question and could be categorized into the available response options and whether misclassifications could be minimized. This initial questionnaire development research was conducted between 2001 and 2013 using data collected through the NHIS and NSFG, as well as cognitive interviews (Dahlhamer et al., 2014; Ridolfo et al., 2012; Miller and Ryan, 2011; Miller, 2001). The cognitive interview respondents ranged in age (18–60+ years), ethnicity and race (more than 75% people of color), and sexual orientation and gender identity (more than 40% identified as something other than straight/heterosexual, and 6% indicated their gender was “more complicated” than male or female). This research demonstrated comprehension of the sexual orientation identity measure among sexual majority and minority populations with racially diverse samples; the item based on the testing was ultimately included in the 2013 NHIS.
Another component of the cognitive standard discussed in Chapter 4 is the extent to which items result in respondents being willing and able to answer them (Suen et al., 2022; Bates, García Trejo, and Vines, 2019; Ellis et al., 2018; Fredriksen-Goldsen and Hudson, 2018; Rullo et al., 2018). The feasibility of asking questions about sexual orientation in population-based data collection has been demonstrated across many studies. For example, Meyer and colleagues (2002) conducted a study that found that women would respond to an anonymous random-digit-dial telephone health survey asking a series of questions that included sexual orientation identity and sexual attraction. More recent studies of feasibility provide further evidence that respondents in population-based surveys and health care settings are willing and able to respond to questions about sexual orientation identity (Bates et al., 2019; Rullo et al., 2018; Ellis et al., 2018; Lee et al., 2018). Several of these studies noted that while there are subsets of respondents who find the questions about sexual orientation sensitive, such concerns are not limited to the construct of sexual orientation: for example, questions
about disability status were also reported as potentially sensitive and yet feasible to collect.
In current surveys, the nonresponse rate for questions on sexual orientation identity appears more than acceptable. There is no one standard for acceptable item nonresponse rates; however, the U.S. Census Bureau, in a review of changes in responses to specific items between population counts, characterized nonresponse under 5.95 percent as “low.”2 The percentage of respondents in 2020 who refused or did not provide data for sexual orientation on the English versions of the NHIS, Behavioral Risk Factor Surveillance System (BRFSS), Census Pulse Survey, and other probability national samples asking this question were all under 3 percent: see Table 5-1 (also see Bates, García Trejo, and Vines, 2019; Meyer et al., 2019). That is, nonresponse rates are lower than or comparable to other demographic or outcome variables, such as race and family income. The point estimates for the proportion of the U.S. population that is identified as a sexual minority does vary across the surveys that use this measure, but without confidence intervals for the point estimates it is not clear whether these differences are significant or meaningful.
The panel also considered mode of administration. The recommended measure has been used in self-reported surveys and interviewer-administered surveys. A study by Dahlhamer and colleagues (2014) found that the accuracy of the measure in national surveys was not statistically different if respondents completed the measure using an audio computer-assisted self-interviewing mode or a computer-assisted personal interview mode. Household surveys often require proxy reporting on household members. A recent effort to test whether a similar measure of sexual orientation identity could be reliably collected by proxy has concluded that it is feasible (Holzberg et al., 2019). However, given the challenges associated with proxy reporting, the panel concluded that that we did not yet have sufficient evidence to endorse the use of proxy reporting.
Comparison with Other English-Speaking Countries
The panel also reviewed standards for collecting data on sexual orientation identity from other English-speaking countries. Outside the United States, Australia, Canada, New Zealand, and the countries of the United Kingdom have begun or recently completed the process of revising their
___________________
2 See https://www.census.gov/newsroom/blogs/random-samplings/2021/08/2020-censusoperational-quality-metrics-item-nonresponse-rates.html.
TABLE 5-1 Item Nonresponse Rates and Sexual Minority Population Prevalence for Recommended Sexual Orientation Identity Measure
Survey | Question | Nonresponse/Unknown Rate | Percentage Lesbian, Gay, Bisexual |
---|---|---|---|
2016 NCVS |
Which of the following best represents how you think of yourself?
|
2.8% | 1.9% (weighted) |
2018 GSS |
Which of the following best describes you?
|
1.9% | 5.8% (weighted) |
2020 BRFSS |
Which of the following best represents how you think of yourself?
|
1.9% | 4.2% (weighted) |
2020 NHIS |
Do you think of yourself as:
|
2.6% | 3.1% (unweighted) |
2021 Census Pulse |
Which of the following best represents how you think of yourself?
|
2.2% | 6.5% (unweighted) |
NOTE: All data from national samples of U.S. adults. Nonresponse rates calculated by committee members from publicly accessible codebooks for Behavioral Risk Factor Surveillance System (BRFSS) and National Health Interview Survey (NHIS); online data tools for the General Social Survey (GSS); and detailed data tables for the U.S. Census Pulse Survey.
SOURCES: Data from BRFSS (CDC, 2021), NHIS (National Center for Health Statistics, 2021), National Crime Victimization Survey (NCVS; Truman et al., 2019), GSS (Davern et al., 2021), and U.S. Census Pulse Survey (U.S. Census Bureau, 2021).
data collection standards for measuring sexual orientation (see Table 5A-2 in the annex to this chapter). All of the countries defined the concept of sexual orientation as an umbrella term that encompasses the dimensions of identity, attraction, and behavior, and they all stress that these characteristics can change over time and do not necessarily correspond to one another at any given time. These countries also use very similar measures for sexual orientation identity: each country offers the same three major response categories, and all include the option of a write-in response.
A question on sexual orientation identity was included for the first time in the 2021 census for England and Wales and will be included in the 2022 census of Scotland. The Office of National Statistics (ONS) conducted extensive pretesting for the measure and ultimately made it a voluntary response for people ages 16 and up (Office of National Statistics, 2021). The format of the question, “Which of the following best describes your sexual orientation?” is consistent across all three territories (see Table 5A-2);3 the major difference between this format and most versions in the United States is that the question stem specifies the term sexual orientation, and a write-in line is provided to account for responses in addition to “Straight or heterosexual,” “Gay or lesbian,” and “Bisexual.” In addition to the censuses, ONS has also been collecting sexual orientation identity on its Annual Population Survey and Crime Survey for England and Wales. Its original 2009 guidance for measuring sexual identity was among the earliest national publications of its kind; the national harmonized standard was last updated in 2016 (Office of National Statistics, 2016).
Canada did not include a question about sexual orientation in its 2021 census but has asked about sexual orientation identity on the Canadian Community Health Survey since 2003 and on its General Social Survey on
___________________
3 See revised Scotland census form here: https://www.scotlandscensus.gov.uk/documents/scotland-s-census-2022-question-set/.
victimization since 2004. Questions on sexual behavior and sexual attraction were added to Canada’s national health survey of adults and youth, in 2015 and 2016, respectively (Statistics Canada, 2020). A proposed national statistical standard for sexual diversity was published in January 2021,4 but has not yet been implemented. Its currently recommended measure for sexual orientation identity is a close version of the United Kingdom’s question stem and response options (see Table 5A-2).
Overall, the panel found it notable that the language used to describe major sexual orientation identities are relatively consistent across countries. There is some divergence in how to label the write-in option (ranging from New Zealand’s “Other, please state” to Australia’s “I use a different term”), and New Zealand’s testing suggested that respondents in that country had “variable understanding” of the terms sexual orientation and sexual identity; as a result, New Zealand opted not to include that wording in its recommended question stem (Stats NZ, 2019: 11; Australian Bureau of Statistics, 2021). However, the general consistency of approaches to measuring sexual orientation identity provide additional evidence that supports our conclusions and recommendations for the United States.
In our review of measures in other English-speaking countries, the panel also noted that the consensus minimum age recommendation for asking about sexual orientation identity across most of the countries is 15 (see, e.g., Australian Bureau of Statistics, 2021); the United Kingdom opted for 16 and over because that is their age of consent and minimum age at marriage (Office of National Statistics, 2021). However, the panel found insufficient evidence to support a minimum age recommendation for the United States. We return to this point at the end of the chapter in our recommendations for future research.
CONCLUSIONS AND RECOMMENDATIONS
Sexual orientation identity has been assessed as a demographic variable in population-based national and state surveys and numerous community studies for decades (Sexual Minority Assessment Research Team, 2009). The panel prioritized a measure that would be easiest to implement across population-based, clinical, and administrative data collection efforts. The panel approached the evaluation and recommendation of sexual orientation measures following the principles delineated in Chapter 2 and the criteria outlined in Chapter 4. With regard to evaluation criteria, the panel relied heavily on the measure’s demonstrated use and performance in both the general population and LGBTQI+ populations; evidence of conceptual grounding, comprehension, and usability; and record of use as a single
___________________
4 See https://www.statcan.gc.ca/en/concepts/consult-variables/gender.
response to enable enumeration across multiple modes of data collection (e.g., interviewer or computer-assisted interviews, administrative forms, etc.).
Our findings are summarized in Table 5-2, following our recommended measure, and then we provide a discussion of how the criteria apply to the recommended measure.
RECOMMENDATION 2: The panel recommends that National Institutes of Health use the following question for assessing sexual orientation identity:
Which of the following best represents how you think of yourself? [Select ONE]:
- Lesbian or gay;
- Straight, that is, not gay or lesbian;
- Bisexual;
- [If respondent is AIAN:] Two-Spirit
-
I use a different term [free-text]
(Don’t know)
(Prefer not to answer)
Question Stem
The stem of the recommended measure does not require knowledge of the term “sexual orientation” and yet frames the question as an issue of identity or how people see themselves, reflecting the cognitive dimension of sexual orientation. The response options all reflect labels used to describe sexual orientation identity, rather than attraction or behavior. The recommended measure allows for respondents to answer their sexual orientation identity using one of these popular terms or to write-in a response for other terms.
Response Categories
In considering the ordering of response categories, the panel decided to list the response options as they appear in existing surveys. The panel considered whether response options could be reordered based on population prevalence, with the most commonly selected category (“straight, that is, not gay or lesbian”) presented to respondents first; however, the panel decided against this approach given insufficient evidence to support it.
Testing the effects of reordering is particularly important because the “straight” category is phrased as a negation of the “gay or lesbian”
TABLE 5-2 Summary of Findings on Sexual Orientation Identity Measure and Evaluation Criteria
Question Stem | Question Responses | Evaluation Criteria | Evaluation |
---|---|---|---|
Which of the following best represents how you think of yourself? |
(Don’t know) (Prefer not to answer) |
Previous use in population-based data collection |
|
Conceptual fit |
|
||
Testing: comprehension and validity |
|
Populations included in testing |
|
||
Testing: errors and nonresponse |
|
||
Adjustments to previously tested item included in recommended measure |
|
||
Weaknesses and challenges |
|
category, and so, changing their order could affect respondent comprehension. The panel also considered the necessity of still including the clarification term under the heterosexual category (i.e., “straight, that is, not gay or lesbian”). This phrasing places “straight” in contrast with “gay or lesbian,” which (1) reinforces “straight” as the normative, unmarked category and (2) in combination with “gay or lesbian” logically appears to include all possible responses (“gay or lesbian” vs. “not gay or lesbian”), but it does not because, for example, bisexual is excluded. Inclusion of the “not gay or lesbian” qualifier was prompted from research a decade ago for the NHIS (Ridolfo, Miller, and Maitland, 2012). Cognitive interviews found that many straight respondents had little need to express a salient sexual orientation identity because they represent the heteronormative majority. Instead, these respondents often disidentified from a gay identity, sometimes referred to as a “not-me” identity (McCall, 2003). For these respondents, identifying as heterosexual (as some described it “not gay” or “normal”) served to distance themselves from what they perceived as a stigmatized sexual orientation.
The panel recognizes that the wording of the response category for the heterosexual population is informed by psychometric work that was carried out over the past three decades when social awareness of sexual and gender minority identities was less well understood (Fredriksen-Goldsen and Kim, 2015; Sell, 1997). Evidence from testing this question in recent years, however, suggests that heterosexual respondents may not answer the question correctly without the clarification term. For example, researchers (e.g., Michaels et al., 2017) have noted that among people whose primary language is Spanish, changing the phrasing of the straight option to read the negation portion first (e.g., “not gay [or lesbian], that is, heterosexual”) may improve response accuracy compared with the recommended response phrasing (e.g., “straight, that is, not gay or lesbian”); however, the panel did not have sufficient evidence of testing this change in phrasing in other populations to recommend this change. Similarly, there was insufficient empirical evidence that would allow us to determine whether the performance of the measure would be affected if the clarification portion of this response option was removed. Finally, the qualifier is currently used in several of the largest national surveys, including the BRFSS, All of Us, the National Crime Victimization Survey (NCVS), NHANES, and NHIS (starting in 2022; see Table 5A-1). The panel recommends that the clarification included in the heterosexual answer category (i.e., “that is, not gay or lesbian”) and its wording order be further studied to ensure sufficient understanding among the general population, including those for whom English is not their primary language.
The list of sexual orientation identity labels in the United States is continuously expanding (e.g., see, Goldberg et al., 2020; Rothblum et al., 2020) as part of the assessment of available measures, and the panel weighed the opportunity costs associated with including a more expansive list of sexual orientation identity categories. For example, the panel discussed whether additional response categories need to be added to the recommended measure to capture the use of sexual orientation identities such as “queer,” “pansexual,” or “same gender loving.” However, the panel did not have sufficient evidence to warrant making such a recommendation at this time. Almost none of the general population survey questions the committee considered included these response categories (see Table 5A-1) and none demonstrated testing for comprehension in the general population. Moreover, there is reason to believe that these terms might be understood or accepted differently among sexual and gender minority people, as well as the general population. For example, the panel noted that the term “queer” has different connotations by age: queer is experienced as a slur by older people while younger people have sought to reclaim the term; moreover, it can be used to refer to both sexuality and gender (see, e.g., Flatt et al., 2022; Barsigian et al., 2020).
All of Us included “queer” and “pansexual” as response options in a follow-up question administered to those who selected “None of these and I would like to see more options.” However, even when provided this more extensive list of response options (shown in Table 5A-1), only about half of these respondents selected an identity from the list. Similarly, Meyer and colleagues (2019) conducted an experiment with approximately 9,700 adults age 18 or older, randomly assigning respondents to one of two sexual orientation question wordings. For those who responded “something else,” half of the sample got a follow-up question that included “queer,” “pansexual,” “asexual,” “demisexual,” “same gender loving,” and “none of the above.” However, only 0.6 percent of respondents identified as one of these sexual orientations. The study concluded that most respondents (53 percent) who initially identified as “something else” subsequently identified as “none of the above.”
The panel decided on the use of a write-in category as a strategy that allowed respondents who did not find the existing response categories captured their identity to respond. The panel considered the likelihood that additional categories would need to be collapsed into broader categories or dropped from analyses entirely if an insufficient number of respondents selected a given term. The panel strongly encourages the reporting of the use of these write-in categories in published tabulations of responses. An increase in the use of write-in responses could indicate that the existing response categories are no longer representative of those used in the general population and could indicate a need to revisit the recommended response
categories.5 The write-in identity categories could assist in assessing which sexual orientation identities have become sufficiently prevalent in the general population to target for testing as new response categories.
Considerations for Indigenous Populations
Recognizing the importance of representation and visibility of the U.S. Indigenous population and in accordance with federal mandates for data collection on American Indian and Alaska Native (AIAN) populations and recommendation of the National Congress of American Indians and the Indian Health Service 2020 Strategic Vision and Action Plan (Indian Health Service, 2020), the panel considered including an additional response category for surveys that include a population-based, representative, or Indigenous-community-based sample of AIAN respondents. In weighing the inclusion of this category, the panel was concerned whether misreporting by non-AIAN people as Two-Spirit would preclude accurate representation of Indigenous Two-Spirit populations. To assess the potential impact of “false positive” reports as Two-Spirit among non-Indigenous populations, the panel reviewed data from the All of Us study,6 which included Two-Spirit as a sexual orientation response option. These data showed that more than two-thirds of those who identified as Two-Spirit were not AIAN. Given the high probability of producing data on Two-Spirit people that inaccurately reflects AIAN populations, the panel does not recommend making this a response option for all respondents.
However, inclusion of a Two-Spirit response option for AIAN respondents would provide a way for surveys to explicitly include a category that represents sexual and gender minority populations that are culturally specific to the AIAN population. The need for such a response option is demonstrated by the AIAN LGBTQI+ Two-Spirit Study, funded by the National Institutes of Health (HONOR Project; Cassels et al., 2010)—the only nationally representative study of AIAN LGBTQ health. When AIAN respondents were asked, “Which one category best describes your sexual orientation now?” the responses were are follows: 16 percent Two-Spirit, 30 percent gay, 29 percent bisexual, 15 percent lesbian, 7 percent heterosexual, 2 percent reported their specific tribal word for their sexual orientation, and 1 percent reported “other.” Thus, one of six AIAN respondents
___________________
5 Analysts cannot not assume that all write-in responses for “I use a different term” represent sexual minority identities. Write-ins will need to be examined to determine whether responses represent sexual minority identities and are consistent with existing response categories (and can be recoded accordingly) or need to be treated as (uncodable) missing data. The latter may include “protest” write-ins or other off-topic responses (see Bates, García Trejo, and Vines, 2019).
preferred the Two-Spirit term for sexual orientation. This is underscored by the Indian Health Service’s (2020) recommendation that all clinical and research Indian Health Service forms include “Two-Spirit” as a response option for sexual orientation.
Inclusion of the Two-Spirit response category ensures that data will capture the significant portion of AIAN respondents who would otherwise not identify with existing Eurocentric sexual orientation identity response categories. The Indian Health Service recommendation that the Two-Spirit category be included in data collection efforts reflects the fact that this term is used in both rural and reservation environments. The high prevalence of its use within Indigenous populations in combination with the culturally specific role in sexual orientation identity that the term Two-Spirit holds for this population motivated the panel’s decision to include a Two-Spirit option for AIAN respondents.
The panel recommends that Two-Spirit be included as a response category for participants who self-identify as AIAN in computer-assisted surveys where the response options may be tailored through an algorithm. In such cases, data collection would also need to include a measure of racial or ethnic identity that offers AIAN as a response option and is asked prior to questions on sexual orientation identity. In an interviewer-assisted or paper-and-pencil survey in which response options cannot be tailored, the Two-Spirit responses can be tallied through the open text response option. Although offering different response options based on previous survey responses is not standard survey practice, it is not unprecedented in this context, where response options have often differed by gender of the respondent. The panel recommends this measurement strategy as a way to reduce misappropriation of an identity that has a very particular meaning in AIAN communities, a population with language describing sexual and gender minority status or identities that predate the terms that predominate in the general population. This option allows AIAN respondents to see their culturally specific sexual identities represented, even in general population surveys, while preserving the autonomy of all respondents to write-in “Two-Spirit” or a tribal-specific term in the open-text field. When automated programming of response options is not feasible, researchers need to be cognizant of possibility of the appropriation of an Indigenous identity category by non-Indigenous people in their analyses and interpretation of results.
RECOMMENDED RESEARCH AREAS
The panel’s recommended sexual orientation identity measure is the result of a review of currently available evidence. We concluded there is sufficient evidence to support this approach to asking about sexual orientation
identity in large-scale data collection settings, particularly for general population enumeration and research among English-speaking adults. Additionally, there is evidence of acceptable performance in health contexts, including medical settings, clinical trials, and public health surveillance. We also identified potential challenges and limitations to this empirically validated option. Although the sexual orientation identity question used in several federal surveys and electronic health records is the strongest option available, additional research is needed to assess its ongoing use in administrative contexts (see, e.g., Cooper, Wilson, and Choi, 2017), and to respond to concerns expressed by members of sexual minority communities, scholars, and other stakeholders about some aspects of its operationalization to date (see, e.g., Suen et al., 2020). In addition to the issues noted above, these include the following concerns:
- whether concerns about straight people’s comprehension of the sexual orientation identity terms are still warranted, in English and other languages;
- whether changes are needed to the ordering of the response options;
- how best to integrate standardized questions for sexual attraction as a measure of sexual orientation and determine under which conditions it is equally or more useful than identity measures;
- the need for further assessment of item performance across all survey modes, including proxy reporting, in languages other than English, in all major U.S. racial and ethnic populations, and among youth; and
- the need to reevaluate and expand answer options over time, particularly with regard to plurisexual response categories (e.g., queer and pansexual).
In the rest of this section we briefly review these considerations and offer recommendations for future research.
Clarification for the Heterosexual Response Category
The panel had serious concerns about the use of negating and imprecise language in the “heterosexual” response option (“straight, that is, not gay or lesbian”). However, we had insufficient empirical evidence to determine whether the performance of the recommended measure would be affected if the clarification portion of the heterosexual response option was removed. Given increased social awareness of different sexual identities in the general population, the panel recommends that the clarification included in the heterosexual answer category (“that is, not gay or lesbian”) and its wording order be further evaluated to ensure optimal comprehension by the general
public, including those for whom English is not their primary language. Given the evidence suggesting differences in measure performance across different communities, testing the comprehension and performance of the recommended measure across multiple languages and racial, ethnic, and cultural groups is an important area for research.
Ordering of Response Categories
The panel considered the ordering of response categories and whether response options could be reordered based on population prevalence. However, we ultimately decided against this change. To our knowledge, there is no evaluation data that allows us to determine whether reordering the response categories based on their population prevalence would affect the measure’s performance. This testing is particularly important because the “straight” category is phrased as a negation of the “gay or lesbian” category, and so changing the order could affect respondent comprehension. Given the limited evidence on whether response order would affect response rates, the panel encourages research to examine whether response ordering affects the measure’s performance.
Item Performance of “Attraction” in General Population Data Collection
Sexual attraction does not require individuals to identify or conform with a sexual orientation identity, nor does it require individuals to engage in sexual behavior. While the etiologic and historical review of the different conceptualizations and measures used to describe sexual attraction are beyond the scope of this report, the panel acknowledges the need to assess the performance of sexual attraction measures in future research. Specifically, the panel emphasizes that any entity requiring data regarding respondents’ attractions needs to ensure that the measure is independent of LGBQ status or sexual behavior. Moreover, a future measure of attraction needs to acknowledge multiple forms of attraction (e.g., romantic, erotic), include a response for the absence of attraction in response categories (e.g., asexuality, aromantic), and integrate an expansive conceptualization of the gender identities, bodies, and identities to which a person may be sexually attracted. At this time, there are no measures that have been qualitatively or quantitively tested for use in general population data collection. Future research in this area is warranted.
Evaluating Best Practices for Sexual Orientation Measures among Youth
The statement of task instructed the panel to review and recommend measures to assess sexual orientation and gender identity for the U.S. adult population. However, higher proportions of youth and young adults today identify with a sexual and gender minority identity than previously, and this proportion appears to be growing (Jones, 2021). Also, youth and young adults make up a disproportionate percentage of the sexual minority population (Wilson et al., 2021; Meyer et al., 2019). Moreover, youth surveys served as the site of the earliest administration of population-based measures of sexual orientation (Remafedi et al., 1992). Currently, the national Youth Risk Behavior Surveillance System (YRBSS) and many similar state surveys ask questions about sexual orientation, including identity, and several states and localities collect similar data in their administrative systems, such as child welfare, incarceration, and social services (see, e.g., California Legislative Assembly, 2015). Some of the measures used in these administrative settings have undergone testing for comprehension, accuracy, and feasibility (see, e.g., Steiger et al., 2017; Wilson et al., 2016). As such, the need for an assessment of how well the measures perform in population-based and administrative data collection settings is needed to create best practices.
Another dimension of age is determining the youngest age at which sexual orientation questions can be feasibly and reliably asked. Although other English-speaking countries have decided on age standards for asking about sexual orientation, the panel did not find conclusive evidence to support setting age standards, and, as noted above, were tasked with a focus on adults. Ongoing panels and survey research in the United States collect information from middle and high school students (e.g., Add Health; YRBSS) regarding their attractions, behaviors, and identities. Given these findings, the panel believes that youth may comprehend and respond to questions regarding their sexual orientation. However, in the absence of empirical evidence regarding our recommended measure with children and adolescents, the panel recommends further research to appropriately evaluate its performance across both settings (e.g., schools, health care) and age groups.
Inclusion and Potential Expansion of Identity Categories
There are a growing number of sexual orientation identity labels and increasing popularity in adopting some of these labels. Various factors, in addition to popularity of newer terminology, are relevant to how sexual orientation identity measures will perform in the general population, including whether there is evidence that these terms are understood consistently
across the general population. Continual assessment of evolving language and prevalence of sexual orientation identity labels is required. While terms beyond the historical labels of “gay,” “lesbian,” “bisexual,” and “heterosexual or straight” (such as “queer,” “pansexual”) have not been tested for this broad level of comprehension or standardized in federal and national population-based surveys, the panel supports revisiting the sexual orientation identity categories and testing their use in federal and national population-based surveys when there is sufficient evidence to do so. Ongoing testing examination of how additional response categories may affect the performance of the panel’s recommended measure is warranted, as it may offer the necessary evidence to empirically determine whether different responses need to be integrated into population-based surveys and acknowledge the presence of diverse sexual orientations in our society to the public without negatively affecting the currently established measures.
The panel acknowledges the need for continued psychometric testing of the measure to incorporate changes in response category terminology, particularly as the social meaning of sexual orientation identity continues to evolve. The panel also recognizes that for surveys that are conducted in specific communities and age-specific surveys, a more expansive list of identity categories may be warranted. The panel made a distinction between population-based and administrative data collection and the collection of sexual orientation data in community settings. Our recommended measure may not capture the full range of different sexual orientation identities among sexual and gender minorities. While the recommended “I use a different term” may allow respondents to self-report their own identities, these descriptors may be hard to standardize and aggregate. As a result, community-based studies that focus on the diversity in those populations may require additional categories that are reflective of the language, race, ethnicity, culture, geography, and age of respondents.
The addition of a write-in option for “I use a different term” does not preclude the inclusion of additional response options that are more culturally or age tailored for specific populations or community-based LGBTQI+ surveys. In modes of administration in which follow-up prompts may be easily programmed, for example, the panel suggests opportunities that allow respondents to request a second, more detailed set of answer categories as a follow-up question for those who indicate they prefer an identity that does not appear on the list of responses provided. This method is currently being used in several federal studies (e.g., All of Us) and may help streamline the data collection and coding process. Additional research testing whether a second prompt may offer greater accuracy and performance across diverse populations is warranted.
Proxy Reporting
Self-reporting in surveys is preferred and aligns with the panel’s principle to respect identity and autonomy. However, proxy reporting is used in various federal data collection efforts. Before sexual orientation is collected through proxy reporting, additional research is needed to test the accuracy and quality of the data reported; a proxy may not know how other household members self-identify. Testing is also needed on the associated sensitivity regarding confidentiality, that is, disclosing other household members’ sexual orientation identity without their consent. Future evidence from experimental studies and other survey methodology techniques are needed to understand the acceptability of this approach and its accompanying sources of error (e.g., nonresponse rate, misclassification rates).
RECOMMENDATION 3: To further improve the quality and inclusivity of current measures of sexual orientation identity, the National Institutes of Health should fund and conduct research on the following topics:
- Alternate wording for the “straight” response option that performs equally well as, or better than, the existing recommendation for English- and non-English-speaking populations without using language that negates gay or lesbian identities. Such an assessment should include studies on whether the clarification “that is, not gay or lesbian” is still needed to ensure accurate comprehension among heterosexual populations.
- The ordering of response categories, including sorting response categories based on population prevalence.
- Guidelines for measures that capture other dimensions of sexual orientation, including sexual behavior and sexual attraction: in particular, standards for assessing asexual identities should be developed and tested.
- The utility of including sexual orientation identity response options that may be more prevalent in subsets of the LGBTQI+ population, such as “queer” and “questioning” or, in African American communities, “same gender loving.”
- The performance of existing measures and identification of best practices for how to assess sexual minority status among adolescents, including whether sexual orientation identity is the most effective dimension to track (compared with attraction or behavior) when assessing disparities in well-being, the appropriate age at
- which to begin asking sexual orientation questions, best practices for reflecting feelings about being unsure (i.e., questioning), and the range of response options.
- How reporting of sexual orientation identity is affected when reporting is done by proxy, such as when a single household respondent responds on behalf of all household members.
ANNEX:
SEXUAL ORIENTATION IDENTITY MEASURES IN THE UNITED STATES AND OTHER ENGLISH-SPEAKING COUNTRIES
This annex comprises two tables: Table 5A-1 shows all measures used in the United States to measure sexual orientation identity; Table 5A-2 does the same for other English-speaking countries.
TABLE 5A-1 Measures of Sexual Orientation Identity in the United States
Question | Response Options | Source(s) |
---|---|---|
Which of the following best represents how you think of yourself? |
|
All of Us (NIH) |
[If “None of these describe me”:] Are any of these a closer description of how you think of yourself? |
|
|
Do you consider yourself to be heterosexual or straight; homosexual or gay or lesbian; or bisexual? |
|
American National Election Studies (ANES) Women Only |
Question | Response Options | Source(s) |
---|---|---|
Do you consider yourself to be heterosexual or straight or homosexual or gay or bisexual? |
|
American National Election Studies (ANES) Men Only |
Which of the following best represents how you think of yourself? |
|
Behavioral Risk Factor Surveillance System (BRFSS-CDC) National Crime Victimization Survey (NCVS-DOJ) National Health Interview Survey (NHIS-CDC) National Health and Nutrition Examination Survey (NHANES) Men Only |
Which of the following best represents how you think of yourself? |
|
Behavioral Risk Factor Surveillance System (BRFSS-CDC) National Crime Victimization Survey (NCVS-DOJ) National Health Interview Survey (NHIS-CDC) National Health and Nutrition Examination Survey (NHANES) Women Only |
Which one of the following do you consider yourself to be? |
|
Center for Substance Abuse Treatment-Government Performance Results and Modernization Act (SAMHSA-CSAT-GPRA) |
Do you, personally, identify as lesbian, gay, bisexual, or transgender? |
|
Gallup Daily Tracking Poll |
Question | Response Options | Source(s) |
---|---|---|
Which of the following best describes you? |
|
General Social Survey |
Next, we’d like to ask you a question about how you think of yourself. Do you consider yourself to be [lesbian/gay], straight, bisexual or something else? |
|
Health and Retirement Study (HRS) |
Do you think of yourself as straight or heterosexual; as gay, lesbian, or homosexual; or as bisexual? |
|
Health Center Patient Surveys (HCPS-HHS) |
Now I will read a list of terms people sometimes use to describe how they think of themselves. Lesbian or gay, that is, homosexual Straight, that is, heterosexual Bisexual Don’t know, or Another sexual orientation As I read the list again, please say ‘Yes’ when you hear the option that best describes how you think of yourself. |
|
High School Longitudinal Study of 2009 (HSLS:09) CAPI or CATI interview |
Do you think of yourself as: |
|
High School Longitudinal Study of 2009 (HSLS:09) Web-based interview |
Question | Response Options | Source(s) |
---|---|---|
Do you think of yourself as…? |
|
National Adult Tobacco Survey (NATS-NCHS) |
Which of the categories on the card best describes you? |
|
National Epidemiologic Survey of Alcohol and Related Conditions (NESARC) |
Do you consider yourself to be: |
|
National HIV Behavioral Surveillance (NHBS) |
Which one category best describes your SEXUAL ORIENTATION now? |
|
National HONOR Project Study of LGBTQ+-Two Spirit American Indians and Alaska Natives (Cassels et al., 2010). |
Do you consider yourself to be heterosexual or ‘straight’ or bisexual, or homosexual or gay? |
|
National Inmate Study (Bureau of Justice Statistics) |
Question | Response Options | Source(s) |
---|---|---|
Do you think of yourself as lesbian or gay; straight, that is, not gay; bisexual; Something else? |
|
National Intimate Partner and Sexual Violence Survey (NISVS) |
Please choose the description that best fits how you think about yourself: |
|
National Longitudinal Study of Adolescent to Adult Health (Add Health) National Longitudinal Study of Adolescent and Adult Health (Add Health—SOGI-SES Supplement) |
Which one of the following do you consider yourself to be? |
|
National Survey of Drug Use and Health (NSDUH) |
Do you think of yourself as… |
|
National Survey of Family Growth (NSFG) |
Which of the following best represents how you think of yourself? |
|
National Survey of Older Americans Act Participants (NSOAAP) |
Question | Response Options | Source(s) |
---|---|---|
Which of these best fits how you think of yourself? |
|
National Survey of Youth in Custody (NSYC-2) |
Do you consider yourself to be… |
|
Population Assessment of Tobacco and Health (PATH) |
Which of the following best describes you? |
[Additional response options added in 2021]
|
Youth Risk Behavior Surveillance System (YRBSS) |
TABLE 5A-2 Measures of Sexual Orientation Identity in Other English-Speaking Countries
Question | Answer Options | Source |
---|---|---|
How do you describe your sexual orientation? Please [tick/mark/select] one box: |
|
Australian Bureau of Statistics (2021) |
What is your sexual orientation? Would you say you are: |
|
Statistics Canada (2021) |
Which of the following options best describes how you think of yourself? |
|
Stats New Zealand (2019) |
Which of the following best describes your sexual orientation? This question is voluntary. |
|
United Kingdom: 2021 Census of England and Wales; 2022 Census of Scotland (Office of National Statistics, 2016) |