Methods and sampling
How to find and sample lesbian, gay, bisexual, transgender and intersex people for research, and how to ask about sexual orientation and gender identity so people answer honestly.
Probability and non-probability samples
Early research on gay and lesbian people drew on psychiatric patients, prisoners and bar patrons, and those samples fed the idea that homosexuality was an illness. Probability samples, where everyone in the population has a known chance of selection, are the only way to estimate how many people are LGBTQI+ or to compare their health with everyone else's.
Common probability designs: simple random, systematic, probability proportional to size, stratified, cluster, multi-stage and multi-phase. Because sexual and gender minority people are a small share of most populations, surveys often add a screener: a short set of questions to find eligible people before the full interview.
Non-probability samples (community venues, social media, online opt-in panels) can reach many more sexual and gender minority people and suit many questions, but they can't give population estimates. Online opt-in samples also face fraudulent and inattentive respondents, so they need checks for both.
Respondent-driven sampling
Respondent-driven sampling recruits through participants' own networks: each participant invites a few others, and the analysis adjusts for who knows whom. It was built for hidden populations and has been used widely with men who have sex with men and transgender women. Its estimates depend on assumptions that are hard to check, so report them.
Modes
The mode, meaning how a question is asked, changes who answers and how honestly.
- Self-administered (web, paper, computer-assisted self-interviewing): more reporting of same-sex behavior and fewer refusals than when an interviewer asks. Recommended where possible.
- Interviewer-administered (telephone, face-to-face): works, but answers can shift with privacy and the interviewer. Let the respondent answer by number or letter, or hand over a device.
- Address-based and push-to-web designs: many federal and state surveys now sample addresses and invite people online, with mail or phone follow-up.
- Probability-based online panels: recruited by address or phone, then surveyed online. Useful for SOGI research when the panel's sexual and gender minority sample is large enough.
Sample size
The sample needed depends on the precision you want, your budget, how much the outcome varies, and how efficient the design is. Because sexual and gender minority groups are small, a general-population sample of 10,000 may hold only a few hundred people in a group of interest. Before collecting new data, check the Data catalog: pooling years of an existing survey is often cheaper and larger. For working out sample sizes and weights, see Valliant, Dever and Kreuter, Practical Tools for Designing and Weighting Survey Samples (2nd ed., Springer, 2018).
Privacy
Sexual orientation and gender identity answers can harm people if disclosed. Ask them in private, self-administered modes where you can; explain why you ask and how answers are protected; and don't publish small cells that could identify someone. In proxy reporting, where one household member answers for another, the answer may be wrong or may out the person, so avoid it for these questions where possible.
Research on measurement
Studies that test how sexual orientation questions work, each tied to a survey in the Data catalog.
- Add Health: Russell and Consolacion (2003) looked at romantic attractions and relationships, not just identity labels, and linked them to emotional health in US adolescents. Journal of Clinical Child and Adolescent Psychology
- Minnesota Adolescent Health Survey: Remafedi and others (1992) asked 34,706 students about attraction, fantasy, behavior and identity; 10.7% were unsure of their orientation, and attraction was reported more often than behavior or identity. Pediatrics, PubMed 1557267
- Eight school surveys, including NAIAHS and the BC Adolescent Health Survey: Saewyc and others (2004) compared sexual orientation items across eight US and Canadian school surveys from 1986 to 1999, and recommend standard wording, pilot testing, and attraction as the best single item. Journal of Adolescent Health
- Project HOPE International Survey of AIDS-Risk Behaviors: Sell, Wells and Wypij (1995) found that 8 to 12% of adults in the US, UK and France reported same-sex attraction but no same-sex behavior since age 15, so counting behavior alone misses many people. Archives of Sexual Behavior