Matt Jans, Bianca D. M. Wilson, and Jody L. Herman

Sexual orientation and gender identity (SOGI) are two demographic and personal dimensions that have received relatively little attention in large-scale surveys and official statistics, despite recent advances in social, cultural, and legal equality for lesbian, gay, bisexual, and transgender (LGBT) people.

About 10 million adults in the US (4.1%) identify as LGBT, according to polling conducted by Gallup, up almost 2 million from 2012. Notably, LGBT rights and recognition have seen major advances over the past two decades. For example, according to the Pew Research Center on Religion and Public Life, nationwide support of same-sex marriage has continued to increase. It is currently at 55%, passing the percentage opposing in 2012, and continuing to climb since the 2015 legalization of same-sex marriage. Across religious and demographic groups, 63% of Americans say that homosexuality should be accepted, with acceptance rates of 66% among Catholics, 76% among white mainline Protestants, and 80% among religiously unaffiliated people.

Further, polling on bathroom use rights shows a slight majority support for transgender people using public restrooms in accordance with their gender identity, as opposed to their sex assigned at birth.

Less-quantitative evidence of social and cultural progress of the LGBT community is reflected in the number and types of LGBT characters in mainstream popular entertainment. From the shows “Ellen” and “Will and Grace” in the 1990s, which included representations of gays and lesbians, to the expansion of representation to transgender people and LGBT people of color through “Orange is the New Black,” “Transparent,” “Shameless,” “Black-ish,” “Modern Family,” and “Better Things,” it’s clear that positive and realistic portrayals of LGBT people are now more common than in the past.

In addition to the increasing presence of ethnically and socioeconomically diverse LGBT people in fictional shows, multiple documentary and news segments have been devoted to the issues of transgender people specifically, such as Katie Couric’s NatGeo special “Gender Revolution,” HBOs “Vice Special on Trans Youth,” and NBC’s “Special on Transgender Youth.” All of this indicates a meaningful socio-cultural shift.

With this developing recognition and acceptance, it is surprising that more large-scale and official surveys do not measure these important personal and demographic characteristics.

A “Special Population” or “Part of the General Population”? Specification and Measurement Challenges

Perhaps one reason for scant SOGI measurement is the perspective that LGBT people are part of a “special population” to be targeted with sampling methods that are sometimes at odds with the goals of general population surveys (e.g., non-random convenience samples). Because the rate of LGBT identification in the general population is low, it is usually impractical to use general population survey methods for one-time surveys that focus primarily on LGBT respondents. However, surveys with large samples, or that recur regularly and can be combined over time, are often excellent vehicles for sampling the LGBT population.

A second challenge to the broad inclusion of LGBT people in “standardized surveys” is the seeming complexity of deciding how to measure SOGI. LGBT people have always been sampled in general population surveys, but they have been a “hidden population” because the questions needed for them to identify themselves accurately often have not been used. Alfred Kinsey’s mid 20th-century sexuality research may be partly to blame. Kinsey’s method used open-ended sexual history interviews with complex scoring systems to obtain very nuanced measures of sexual identity, attraction, and behavior that are not only infeasible in general population surveys, but have large reliability and reproducibility concerns. However, measuring SOGI need not be so complex.

At the other extreme, and a much better representation of SOGI measurement best practices for surveys, the Gallup statistic cited earlier uses an all-inclusive definition, asking one question that uses common identity labels that have sociopolitical and cultural meaning to most people: “Do you, personally, identify as lesbian, gay, bisexual, or transgender?” Following Gallup, a survey could include SOGI measurement by adding only one question. However, gains in sensitivity are balanced by losses in specificity and the inability to separate the sometimes overlapping subgroups of lesbian, gay, bisexual, and transgender individuals for analysis.

There is general agreement among sexuality researchers that sexual orientation has three primary dimensions: sexual attraction (i.e., whether a person is physically attracted to people of the same sex, other sex, or both), sexual behavior (whether a person has had sexual relationships with people of the same sex, other sex, or both), and sexual identity (the term a person uses to define their sexual orientation, such as gay, lesbian, bisexual, asexual, straight, or something else).

Each measure has its own strengths and limitations. In the mid-1980s, public health researchers began to realize that some men who had sex with men (MSM) did not identify themselves as “gay,” meaning that survey questions using terms such as “gay” or “homosexual” would not accurately capture their orientation. Thus, asking the sex of sexual partners was the right way to identify this sexual minority group.

Of course, sexual behavior measures do not capture the full spectrum of sexual orientation, particularly among people who have not been sexually active during the timeline defined by the survey. In contrast, sexual attraction measures offer the most inclusive definition, and tend to identify the largest number of people who vary from the heteronormative framework, including unexpressed attraction. However, looking across general population surveys, particularly those used for public health surveillance, measures of sexual identity and sexual behavior tend to be the most common, probably due to the direct public health uses of these variables for risk and exposure analyses.

The situation for gender identity measurement is somewhat different. Most surveys already have a measure of sex (e.g., “Are you male or female?”), sometimes referred to as “gender” for numerous reasons. Nuances of labeling and polite terminology aside, such a binary perspective on gender is overly simplistic and fails to capture transgender identification and other gender nonconformity (as well as physical conditions sometimes called “intersex” or “differences of sexual development” [DSD]).

People whose gender identity matches their sex assigned at birth (i.e., for whom the traditional sex question above would be easy to answer) are sometimes called cisgender, while the term transgender (refers to people whose internal perception of their gender does not match the sex they were assigned at birth). Gender nonconforming is sometimes also used by people who feel they do not fit traditional sex-based gender expression roles in some way.

Many more terms describe specific gender identities and expressions, and new terms develop all the time, particularly among younger people.

Nonetheless, survey items about gender identity that include transgender identification have been used in health research, particularly HIV research, for decades. Yet, questions about gender identity have not been part of large-scale surveys until relatively recently and are only now, slowly, becoming a common part of the dialogue for survey planning.

How Many Surveys Measure SOGI

The Federal Interagency Working Group on Improving SOGI in Federal Surveys of the U.S. Office of Management and Budget (which sets statistical policy for U.S. federal statistics) recently conducted a review and summary of the federal surveys that measure SO and GI. Across all federal surveys and studies, 11 surveys (and one study) included some measure of SOGI, most within the Department of Health and Human Services. While each of the 11 surveys included a measure of self-identified sexual orientation, three also collected sexual attraction and four collected sexual behavior.

Gender identity, on the other hand, was measured much less often, with only six of these 12 surveys/studies (50%) measuring gender identity (in perspective, 100% of the surveys asked about self-identified SO, 25% also asked about sexual attraction, and 33% asked about sexual behavior).

Some of the most-interesting developments in the U.S. federal statistical system are in the Behavioral Risk Factor Surveillance System (BRFSS), which has included an official optional SOGI module since 2014, but also allows individual states to include their own SOGI measures. The first state to ask about transgender status and sexual orientation in BRFSS was Vermont in 2000, although this question’s wording was quite inaccurate and insensitive, focusing on cross-dressing and using the term “transvestite,” which do not necessarily indicate a transgender identity. The question has since been changed to a more-inclusive and -accurate terminology).

The BRFSS as a whole is a pioneer in the formalization of SOGI measurement in population surveys. With its range of questions and national scope, it offers an interesting opportunity to look at the effect of question wording on regional variation in identification rates.

While federal surveys are some of the largest and most-used general population surveys, they are not the only ones to ask SOGI questions. The General Social Survey (GSS) also has a long history in SO measurement and has asked about sexual behavior (sex of sex partners) since 1988 and a sexual identity question since 2008. The California Health Interview Survey (CHIS) has asked about sexual identity and sexual behavior since 2001, and GI since 2015. Notably, many surveys only ask these questions of adults, although CHIS now asks gender expression of teens. (For other surveys and studies that have measured SOGI, see the Further Reading recommendations.)

General Advice and Best Practices for Measuring SOGI and Sampling the LGBT Population

The OMB reports discussed above focus on current U.S. federal statistical practice and future directions for including SOGI measures in officials surveys. For readers seeking best practices in SOGI measurement, we recommend reviewing the SMART and GenIUSS reports published by the Williams Institute. Drawing on those and our related experience in testing and implementing SOGI questions, we provide some general best practices, followed by questions here and answers about SOGI measurement.

1) Clearly define the LBGT population to measure within the target population of the survey and goals of the study: Most “general population” surveys begin with a definition of the target population to sample. By definition, these surveys will include LGBT and non-LGBT respondents. If large numbers of LGBT respondents are required for analyses, over-sampling plans or cross-year combining (if the survey will be repeated) should be considered. Further, for SOGI measurement within general population surveys, researchers and survey administrators should take care to make sure the specific SOGI measures they use are pertinent to the goals of the survey and statistical reporting.

For example, if the point of the survey is studying sexually transmitted diseases, a sexual behavior measure will be more useful than a sexual attraction or sexual identity measure. If the purpose of the study is to document rates of harassment by police among transgender sex workers, gender expression may be more important to measure than gender identification. An investigator may sometimes wish to assess all of these possible dimensions, but every researcher has to consider which dimensions of SO and GI are needed most for the aims of their study.

2) Make sure questions are population- and age-appropriate: Some peoples’ sexual and gender identities are experienced early in life, while others unfold later. Similarly, a person’s sexual experiences tend to shape their identity over time. A youth may be able to indicate with certainty that they are attracted only to members of the same sex, but may not have had the opportunity to express that attraction. Particularly with younger people, it can be better to ask about sexual identity and attraction than sexual behavior.

3) Pre-test measures, even if they are taken from existing surveys or follow best practices: While there are now comprehensive best practices for SOGI measurement, every survey should pre-test selected measures and interviewing procedures with people similar to those who will be in its sample. There is still much to learn about the role of survey mode and question placement in measurement error and item non-response on SOGI questions.

Particularly in the context of general population surveys, the terminology used to measure SOGI must be understood equally by respondents who identify as a sexual/gender minority and those who do not. Thus, questions must be written in relatively plain language, so respondents can answer whether or not they understand SOGI-specific terminology. For example, the Williams Institute and the California Health Interview Survey experimentally tested gender identity questions in an adult population, and found that a two-question measure using simple language performed better than a one-question measure using complex terms. Without that pre-testing, we might have been tempted to select the one-question measure simply because it was a single question. However, our pilot test clearly showed that the two-question measure was the right choice.

4) Consider the data collection context: This includes the mode of data collection, physical context of data collection, and issues such as the age of respondents sampled for the survey. For example, will questions be asked of each member of the household in a fairly open setting where all respondents (and perhaps even those who are not selected to respond) can hear the questions and answers? Will questionnaires be administered in a classroom or school library, where classmates could observe students’ responses (even if the questions are self-administered)?

Such methodological details could seriously affect accuracy of response and missing data rates, and should be addressed conscientiously in the survey design. In general, it is recommended to administer SOGI questions in a self-administered mode, or other private mode, when possible to make the respondent more comfortable when answering, and reduce the risk of disclosure to others present.

Frequently Asked Questions (FAQs) about SOGI Measurement

Question #1: What specific questions should I ask to measure sexual orientation accurately?

According to the Williams Institute’s SMART report, the recommended way to ask about self-identified sexual orientation is, “Do you consider yourself to be heterosexual or straight, gay or lesbian, or bisexual?” Some experts advise against using the technical terms “heterosexual” and “homosexual” because it requires a relatively high level of literacy to understand them. Pre-testing can help address whether this is a concern for the specific population being studied, although including them sometimes facilitates translation into other languages that do not have a term for “straight.”

To measure sexual behavior, the recommended approach is to ask, “In the past [DEFINE TIME PERIOD], who have you had sex with? Men only? Women only? Both men and women? I have not had sex in [DEFINE TIME PERIOD].” While a questionnaire design purist would place a screener question asking about sexual activity in the defined time periods first, then skipping out people who have not had sexual partners in that period (see method used in CHIS), this question is otherwise simple and straightforward. Because it does not ask for number of sexual partners, this question could as easily be answered for the past year, past decade, or lifetime. Modifications would obviously have to be made for interviewer-administered modes.

Finally, the recommended method of measuring sexual attraction is to ask, “People are different in their sexual attraction to other people. Which best describes your feelings? Are you only attracted to females, mostly attracted to females, equally attracted to females and males, mostly attracted to males, only attracted to males, or not sure?” For both attraction and behavior, if possible, the response options should be ordered based on the respondent’s sex so other sex options are asked first (e.g., “attracted to women” would appear first for men).

Question #2: What questions should I ask to measure gender identity accurately?

Following the GenIUSS report about gender identity measurement, and pilot testing conducted on the California Health Interview Survey, a two-step (i.e., two-item) measure appears to be the simplest way to ask about gender identity while having a minimal impact on the overall duration of the survey (compared to a one-step version that provides a definition of transgender and asks respondents whether they consider themselves to be transgender).

Gender Identity Measure Step 1:
What sex were you assigned at birth, as shown on your original birth certificate?
Male
Female
Gender Identity Measure Step 2:
How do you describe yourself? (check one)
Male
Female
Transgender
Do not identify as female, male, or transgender

In interviewer-administered surveys, or when there is space in a self-administered questionnaire, respondents can be allowed to provide their own gender identity terms, allowing for a fully accepting and affirming respondent experience while producing no additional burden for most respondents. Capturing such data on recurring surveys also facilitates tracking the evolution of gender identity and expression terms over time. This can be done for sexual orientation questions as well, and can help give a voice to hidden sexual and gender minorities.

Question #3: Where should I place the questions in the survey, and what mode should I use?

If possible, SOGI questions should be placed in self-administered sections of surveys (e.g., paper and pencil, or computer-based instruments where an interviewer is not present). There is some evidence that asking questions this way results in obtaining more-accurate answers. However, using item non-response as a measure of sensitivity, there does not seem to be a problem with asking SOGI questions in telephone surveys, where the interviewer is not physically present with the respondent.

Specific recommendations can vary depending on the specific measure used. For GI, despite the need for empirical testing, it is generally recommended to place the two-step measure in the same location as the traditional sex question (essentially replacing that single question with the two-step measure). For sexual orientation, the advice is slightly more complex. For general surveys (i.e., where the main topic of the survey isn’t sexuality), sexual identity measures can be placed in demographics sections.

However, if the interview or questionnaire includes a section or questions about sexual behavior or sexual history, then questions can be placed there. One caution is not to place SO questions in a section that is so sensitive or personal that it might experience higher rates of item non-response (e.g., sexual violence), leading respondents to skip the relatively innocuous SO questions.

As with most questionnaire design decisions, context and order of the entire questionnaire must be taken into account. When sexual identity is considered a demographic characteristic, it fits comfortably with questions on age, race, and education. However, questions about the gender or sex of one’s sex partners, or the people to whom one is attracted, probably would seem out of place with typical demographic questions.

Question #4: There seem to be a lot of different terms for different sexual and gender identities. I don’t feel like I know all those terms. I don’t want to offend LGBT respondents by excluding terms or using words that are out-of-date, but also don’t want confuse non-LGBT respondents by including a lot of terms that don’t apply to them. Help!

Remember that in general population surveys, the goal is to write questions that work well for both LGBT and non-LGBT respondents. The goal is to make sure that every respondent has a place to, as comfortably as possible, describe their situation with respect to the question asked. Further, there is much heterogeneity in terminology among the LGBT community, with age, race/ethnicity, and location influencing the terms people use to describe themselves. This means that in most general population surveys, providing lists of LGBT-specific terms isn’t possible. However, when respondents volunteer other terms, they should be recorded.

Particularly if one of the objectives of the study is to capture contemporary terminology, then “other, specify” options can be included. The challenge with these is that they then must be “back coded” into the standardized response categories for reporting. This is particularly important with low-incidence characteristics like SO and GI.

Question #5: My survey doesn’t have any questions about sexuality in it at all, and it seems like such a sensitive topic to add. Won’t respondents find these questions offensive, because they deal with such personal and sensitive topics? Will they abandon my survey?

There is no evidence that SOGI questions obtain particularly high levels of item non-response or induce break-off (unit nonresponse). In fact, income regularly produces the most item nonresponse across surveys, modes, and wordings, and most surveys still include some measure of income in their demographics sections. Drug and alcohol use also have higher levels of item non-response than SOGI measures, and these questions are asked in many surveys.

While there is some evidence that sexual orientation item non-response can be higher in racial/ethnic minorities and respondents with limited English proficiency, this may be due to question wording and translation as much as sensitivity of the topic. It is important to remember that sensitivity in the survey interview context, with the right protections for privacy, can be more like sensitivity in a doctor’s office than sensitivity in the workplace.

Question #6: If plain language is to be used, and questions are supposed to be simple, how can I be sure that all respondents understand terms like “gay” and “transgender” equally? Shouldn’t I use technical and “proper” terms like “homosexual” and “heterosexual” instead of “gay” and “straight”?

Respondents’ understanding of technical terminology is a common problem in survey measurement. While some survey best practices recommend that all respondents receive any definition given, others recommend that interviewers be trained to provide definitions as needed. The most error-prone situations arise when respondents must ask for a definition, because those who do not realize they need the definition to answer accurately will not request it. Thus, the safest approach is to phrase questions as simply as possible to avoid the need for definitions.

Pre-testing, such as cognitive interviewing, can help establish whether the terms chosen are understood equally by a cross-section of the population surveyed. Based on our experience in testing gender identity questions on CHIS, there appears to be no need to provide a detailed definition of transgender; in fact, providing such definitions seems to make a question harder to answer than asking two simple questions (sex assigned at birth and current gender identity).

With respect to using popular language (e.g., gay/straight) versus technical terms (e.g., heterosexual/homosexual), there is good reason to consider using common terminology rather than technical terminology.

In Conclusion

It has never been easier to measure SOGI on large-scale surveys. Successes in multiple federal surveys, state-based large-scale surveys like CHIS, and polling conducted by Gallup are great examples of what is possible. There is no evidence of risk to overall data quality by asking SOGI measures and the quality of the resulting data is substantially enhanced with respect to studying sexual and gender human rights.

Based on social, economic, and health research showing disparities among sexual/gender minorities, it is imperative that more surveys add SOGI measures to ensure that these marginalized populations are a regular part of the data-based decision-making world in which we live.

Further Reading

Dahlhamer, J.M., Galinsky, A.M., Joestl, S.S., and Ward, B.W. 2014. Sexual Orientation in the 2013 National Health Interview Survey: A Quality Assessment (PDF download). 2014. Vital and Health Statistics, Series 2: 169.

Federal Interagency Working Group on Improving Measurement of Sexual Orientation and Gender Identity in Federal Surveys. 2016. Current Measures of Sexual Orientation and Gender Identity in Federal Surveys (PDF download).

Federal Interagency Working Group on Improving Measurement of Sexual Orientation and Gender Identity in Federal Surveys. 2016. Toward a Research Agenda for Measuring Sexual Orientation and Gender Identity in Federal Surveys: Findings, Recommendations, and Next Steps (PDF download).

Gates, G. 2017. In U.S., More Adults Identifying as LGBT. Gallup Social Issues.

GenIUSS Group. 2014. Best Practices for Asking Questions to Identify Transgender and Other Gender Minority Respondents on Population-Based Surveys.

Grant, D., Jans, M., Park, R., Ponce, N., Kil, J., Gates, G., D.M. Wilson, B., and Herman, J.L. 2015. Putting the “T” in LGBT: A Pilot Test of Questions to Identify Transgender People in the California Health Interview Survey (PDF download).

Reisner, S.L., Conron, K.J., Tardiff, L.A., Jarvi, S., Gordon, A.R., and Austin, S.B. 2014. Monitoring the health of transgender and other gender minority populations: Validity of natal sex and gender identity survey items in a U.S. national cohort of young adults. BMC Public Health 14(1):1,224.

Sexual Minority Assessment Team (SMART). 2009. Best Practices for Asking Questions about Sexual Orientation on Surveys.

Wilson, B.D.M., and Miyashita, A. 2016. Sexual and gender diversity within the Black men who have sex with men HIV epidemiological category. Sexuality Research and Social Policy. doi:10.1007/s13178-016-0219-z.

Further Watching

“Gender Revolution: A Journey with Katie Couric,” Nat Geo.

“Trans Youth,” VICE/HBO.

“Transgender Kids,” NBC.

About the Authors

Matt Jans, PhD, is senior survey manager and methodologist with ICF. His PhD is in survey methodology from the University of Michigan. He conducted the work reported here primarily while he was survey methodologist at the California Health Interview Survey at the UCLA Center for Health Policy Research.

Bianca D.M. Wilson, PhD is the Rabbi Barbara Zacky Senior Scholar of Public Policy at the Williams Institute, UCLA School of Law. She holds a PhD in psychology from the Community and Prevention Research program at the University of Illinois at Chicago.

Jody L. Herman, PhD is a scholar of public policy at the Williams Institute, UCLA School of Law. Her PhD is in public policy and public administration from the George Washington University.

Back to Top