Responses to Sexual and Gender Identity Measures in Population-Level Data by Birth Cohort: A Research Note

Abstract

The measurement of sexual and gender identity in the United States has been evolving to generate more precise demographic estimates of the population and a better understanding of health and well-being. Younger cohorts of sexual- and gender-diverse adults are endorsing identities outside of the lesbian, gay, bisexual, and transgender (LGBT) labels. Current population-level surveys often include a category such as “something else” without providing further details, and doing so inadequately captures these diverse identities. In this research note, our analysis of the most recent federal data source to incorporate sexual and gender identity measures—the Household Pulse Survey—reveals that younger birth cohorts are more likely to select “something else” for their sexual identity and “none of these” for their gender identity. The observed sexual and gender identity response patterns across birth cohorts underscore the importance of developing and applying new strategies to directly measure sexual- and gender-diverse adults who identify with identities outside of those explicitly captured on surveys. The integration of sexual and gender identity measures in population-level surveys carries broader implications for civil rights and for addressing health inequities and therefore must be responsive to cohort differences in identification.

Introduction

One goal of President Biden’s Executive Order 14075 on Advancing Equality for Lesbian, Gay, Bisexual, Transgender, Queer, and Intersex Individuals (Exec. Order No. 14075, 2022), issued on June 15th, 2022, was to strengthen federal sexual orientation and gender identity (SOGI) data collection to advance equity for LGBTQIA+ individuals. The LGBTQ Data Inclusion Act (2022) has been proposed in Congress to improve the federal data infrastructure, and federal interagency working groups have assembled to improve and innovate on SOGI measurement in federal surveys (Federal Interagency Working Group 2016a, 2016b, 2016c). The exploration and integration of SOGI measurement strategies have significant civil rights and policy implications (Gates 2013; Gates and Herman 2018). Federal agencies can use SOGI measures in the pursuance of civil rights and equal employment enforcement. The stated goal of the U.S. federal government’s Healthy People 2030 initiative, to address health disparities based on sexual and gender identity, will require SOGI measures to track progress. Further, given the changing political climate at the state level regarding hostility toward LGBTQ+ individuals (ACLU 2023; Movement Advancement Project n.d.), appropriate population-level data sources are needed to examine the impact of differing state-level contexts for the LGBTQ+ population.

Advancement in the federal SOGI data infrastructure is necessitated by the growth in individuals’ LGBT identification—recent Gallup estimates indicate that 7.2% of U.S. adults in 2022 identified as LGBT (Jones 2023). The growth in LGBT identification has occurred alongside a shifting social climate toward LGBTQ+ rights (Gallup, n.d.). The highest prevalence of LGBT identification is found among those born between 1997 and 2003, among whom nearly one in five identify as LGBT (Jones 2023). Younger cohorts of sexual- and gender-diverse individuals also embrace a broader spectrum of sexual and gender identities in addition to, or outside of, the explicitly captured LGBT survey categories. Examples of these other identities include pansexual, queer, gender-nonconforming, gender-fluid, and nonbinary (Brown 2022; Goldberg et al. 2020; Lagos 2022; Morandini et al. 2017; Watson et al. 2020).

Surveys often incorporate identities outside of LGBT into a catch-all category, such as “something else,” which hinders direct examination because it lacks additional information about the respondents’ rationale for their response, and not all individuals who fall into this category can be assumed to be a part of the LGBTQ+ population (Eliason and Streed 2017; Ridolfo et al. 2012; Truman et al. 2019). In addition to being selected by respondents who use sexual or gender identity labels outside those provided, these response categories can also be selected by respondents who do not comprehend the question (Ridolfo et al. 2012; Truman et al. 2019). Rather than representing a definitive category to directly identify sexual- and gender-diverse individuals who use different identity labels, “something else” responses in the absence of additional information function essentially as an uninterpretable “residual” category. Therefore, it is important to ensure that these measurement limitations do not disproportionately disadvantage younger cohorts of sexual- and gender-diverse adults.

In this research note, we investigate variation in responses of “something else” and “none of these” to SOGI measures across birth cohorts using recently collected population-level data from the Household Pulse Survey (HPS). In examining responses to sexual identity and gender identity measures according to birth cohort, we argue that SOGI measurement approaches must be attuned to changing patterns in how people self-identify and underscore the importance of establishing more direct methods to identify sexual- and gender-diverse individuals who identify outside of the LGBT label.

Federal SOGI Data Infrastructure

A recent consensus report by the National Academies of Sciences, Engineering, and Medicine (NASEM) (Bates et al. 2022) offered recommendations for improved practices within the federal SOGI data infrastructure (see Table 1). To address the inability to directly identify sexual- and gender-diverse individuals who embrace identities outside of LGBT the committee recommended a supplemental text response for those who select “I use a different term” for their sexual or gender identity (Bates et al. 2022). The report underscored that “I use a different term” is preferable to “something else,” as the latter language can have negative connotations (Bates et al. 2022).

Table 1.

SOGI measurement strategies

Survey	Year	Sexual Identity	Gender Identity
National Academies of Sciences, Engineering, and Medicine committee recommendations	2022	“Which of the following best represents how you think of yourself?” 1. Lesbian or gay 2. Straight, that is, not gay or lesbian 3. Bisexual 4. [If respondent is American Indian or Alaska Native] Two-Spirit 5. I use a different term [free text] 6. Don’t know 7. Prefer not to answer	“What sex were you assigned at birth, on your original birth certificate?” 1. Female 2. Male 3. Don’t know 4. Prefer not to answer “What is your current gender?” 1. Female 2. Male 3. Transgender 4. [If respondent is American Indian or Alaska Native] Two-Spirit 5. I use a different term [free text] 6. Don’t know 7. Prefer not to answer
American Community Survey	2024 proposed testing	“Which of the following best represents how <Name> thinks of themselves?” 1. Gay or lesbian 2. Straight—that is not gay or lesbian 3. Bisexual 4. This person uses a different term [free text]	“What sex was <Name> assigned at birth?” 1. Male 2. Female “What is <Name’s> current gender?” 1. Male 2. Female 3. Transgender 4. Nonbinary 5. This person uses a different term [free text]
National Health Interview Survey	2022	“Do you think of yourself as:” 1. Gay/lesbian 2. Straight, that is, not gay/lesbian 3. Bisexual 4. Something else 5. I don’t know the answer	“What sex were you assigned at birth, on your original birth certificate?” 1. Male 2. Female 3. Refused 4. Don’t know Experimenting with two gender measures: “Do you currently describe yourself as a man, as a woman, or in some other way?” 1. Man 2. Woman 3. Some other way [free text] 4. Refused 5. Don’t know “How would you describe yourself? 1. Male 2. Female 3. Transgender 4. None of these [free text] 5. Refused 6. Don’t know
Behavioral Risk Factor Surveillance System	2022	“Which of the following best represents how you think of yourself?” 1. Lesbian or gay 2. Straight, that is, not gay 3. Bisexual 4. Something else 5. I don’t know the answer 6. Refused	What was your sex at birth?” 1. Male 2. Female 3. Don’t know/not sure 4. Refused “Do you consider yourself to be transgender?” 1. Yes, Transgender, male-to-female 2. Yes, Transgender, female-to male 3. Yes, Transgender, gender nonconforming 4. No 5. Don’t know/not sure 6. Refused
Household Pulse Survey	2022	“Which of the following best represents how you think of yourself?” 1. Gay or lesbian 2. Straight, that is not gay or lesbian 3. Bisexual 4. Something else 5. I don’t know	“What sex were you assigned at birth, on your original birth certificate?” 1. Male 2. Female “Do you currently describe yourself as male, female, or transgender?” 1. Male 2. Female 3. Transgender 4. None of these

In the U.S. population-level data infrastructure, the federal surveys often used in demographic estimates of the population—the American Community Survey (ACS) and the Current Population Survey (CPS)—have yet to formally include SOGI measures (Manning et al. 2022). Instead, examinations with the ACS and CPS are restricted to same-sex couples, a method that underestimates LGBTQ+ individuals (Deng and Watson 2023). Recently, the Census Bureau proposed the 2024 ACS SOGI Test (Agency Information Collection Activities 2023), and the proposed questions are provided in Table 1. For now, to study sexual- and gender-diverse populations, researchers generally rely on two other federal population surveys—the National Health Interview Survey (NHIS) and the Behavioral Risk Factor Surveillance System (BRFSS) (e.g., Hsieh and Liu 2019; Lagos 2018; Liu and Reczek 2021). More recently, the U.S. Census Bureau launched an experimental data product, the Household Pulse Survey, which in July 2021 integrated SOGI questions (Anderson et al. 2021).

As shown in Table 1, in measuring sexual identities, the NHIS and BRFSS include “straight,” “lesbian/gay,” bisexual,” “something else,” and “don’t know” response categories. Neither the NHIS nor the BRFSS include a supplemental text-response option for those who select “something else” for their sexual identity. For gender identity, the 2022 NHIS experimented with two different measures (National Center for Health Statistics (NCHS) 2023). Respondents in the NHIS were presented with a two-question procedure, which aligns with current recommended practices (Bates et al. 2022; GenIUSS Group 2014; Tate et al. 2013). Respondents reported their sex assigned at birth with binary responses of “male” and “female.” Then, they answered one of two experimental questions. The first asked respondents how they currently described themselves, with options of “male,” female,” “transgender,” or “none of these.” The second provided options of “man,” “woman,” or “some other way.” Respondents who selected “none of these” or “some other way” were provided with an additional text-response option. In the BRFSS, respondents were asked if they considered themselves to be transgender in a one-step question, with options of “yes, transgender, male-female,” “yes, transgender, female-male,” “yes, transgender, gender nonconforming,” and “no.”

For this research note’s goals, the NHIS and BRFSS are not ideal data sources for our examination of responses by birth cohort. When using the NHIS, researchers must combine multiple waves to generate statistically robust sample sizes of nonstraight respondents (e.g., Hsieh and Liu 2019; Liu and Reczek 2021). Considering the relatively recent incorporation of gender identity in 2022 to the NHIS (NCHS 2023), such pooling is not feasible. However, even for sexual identity responses, such analyses may be problematic. Pooling waves of the NHIS would encompass a period estimate of birth cohort responses to the sexual identity measures, and as an average estimate would encompass possible shifts in identification that may have occurred over the pooled years (Gates 2013). It would be ideal to attain a contemporary depiction of SOGI responses in as short a range of pooled years as possible (Gates 2013). Moreover, the NCHS (2020a) cautioned against analyses comparing estimates using pooled waves of the NHIS that contained the years 2019 and onward with those prior to 2019, given the considerable changes that occurred in 2019 in the calculation of sampling weights and the redesigning of the questionnaire and data collection (see NCHS 2020b). For the current study, more recent birth cohort years of 2001–2004 would be observed only in the post-redesign years, whereas older cohorts would appear in both pre-redesign and post-redesign years.

Moreover, although the BRFSS includes both sexual and gender identity measures, not all states have opted into the SOGI module that includes the questions, and the states with available data have fluctuated over the years (National Center LGBT Cancer Network n.d.). To adjust estimates, researchers have used methods such as multilevel regression with poststratification (MRP) to account for states that do not employ the SOGI module, in addition to pooling multiple waves together. For instance, using the 2020–2021 BRFSS and MRP, Flores and Conron (2023) estimated the U.S. adult LGBT population at 5.6%. Nonetheless, a limitation of the BRFSS for purposes of the current study is that although the sexual identity question includes a “something else” response, the gender identity measure does not, and this latter response is of central interest.

Given the current constraints of population-level data sources, we relied on the Household Pulse Survey. As shown in Table 1, the HPS offered respondents similar options for sexual identity as the NHIS and BRFSS: “gay or lesbian,” “straight,” “bisexual,” “something else,” and “don’t know.” Additionally, as shown in Table 1, the HPS used a two-step procedure to measure gender identity, first asking respondents their sex assigned at birth, with options of “male” and “female,” and then their gender identity, with options of “male,” “female,” “transgender,” and “none of these.” The HPS also allows us to examine cohort differences in the responses of both “something else,” to the sexual identity question, and “none of these,” to the gender identity question. A key benefit of the HPS sampling strategy is its substantial size, which does not necessitate pooling across years to garner a powered sample, thus overcoming concerns with pooling multiple waves to examine cohort differences in responses (Gates 2013).

Despite its advantages, there are notable caveats of the HPS that warrant consideration when using these data. First, the response rates of the HPS are relatively lower than those of other Census Bureau data products (Julian et al. 2024). Weighting adjustments by the Census Bureau helped mitigate this nonresponse bias; however, it is likely that some biases remain (Peterson et al. 2021). Second, as an experimental data product, the Census Bureau cautioned that the HPS should not be interpreted as the single indicator of LGBT prevalence in the United States (Anderson et al. 2021). Before examining how responses in the HPS vary by birth cohort, we triangulate experimental SOGI estimates from the HPS with other data sources. Doing this will inform how estimates from HPS compare to those from other sources that are more commonly used for SOGI estimates and contextualize the subsequent analyses.

Triangulating Data Sources

Directly triangulating estimates for SOGI responses is difficult in the current data infrastructure because of differences in the mode of administration, population coverage, sample sizes, and question/response construction. The HPS is fully conducted online (U.S. Census Bureau 2022), the BRFSS is conducted through landline and cell phone interviews (National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP) 2023a), and the NHIS is conducted via in-person interviews (NCHS 2023). Other reports have speculated that differing modes of interviews may contribute to different estimates of sexual identity, with online interviewing providing more privacy (Deng and Watson 2023).

While the HPS and NHIS can generate estimates representative of all fifty states (NCHS 2023; U.S. Census Bureau 2022), unadjusted estimates from the BRFSS are generalizable to the states that have opted into the SOGI module. Although the HPS includes measures for state of residence (e.g., Anderson et al. 2021), enabling direct comparison of BRFSS states with the HPS, the public-use NHIS files cannot provide respondents’ state of residence and can be accessed only via the secure Research Data Center (NCHS 2023), limiting direct comparisons of the NHIS and BRFSS states that have adopted the BRFSS SOGI measures. Although, as shown in Table 1, the NHIS, BRFSS, and HPS adopted similar sexual identity measures, the gender identity measures incorporated are not directly comparable. The NHIS is still experimenting with two measures that are accessible only through the Research Data Center (NCHS 2023), and the BRFSS and HPS use starkly different methods to measure gender identity.

We triangulate the data to determine how estimates from the HPS may differ from those of other datasets, and we weight estimates from each data source as directed in the respective user manuals (NCCDPHP 2023a; NCHS 2023; U.S. Census Bureau 2022). In Figure 1, we present the 2022 estimates for sexual identity using the HPS, NHIS, and BRFSS. Estimates from the sexual identity responses are provided in Panel 1, while all nonstraight sexual identity responses are presented with 95% confidence intervals in Panel 2. To facilitate direct comparisons between the 2022 BRFSS and HPS, we limited the BRFSS to the 33 states that incorporated the SOGI module that year (NCCDPHP 2023b). For the HPS, we restricted our estimates to these same states, which are referred to as “Select BRFSS” and “Select HPS” in Figure 1. Although in 2022 the BRFSS included Guam (NCCDPHP 2023b), the HPS does not generate estimates for Guam (U.S. Census Bureau 2022), thus we removed it for our select comparison. The “U.S. HPS” and “U.S. NHIS” estimates provided in both panels include estimates that are representative of the 50 states and the District of Columbia.

Fig. 1.

2022 adult sexual identity estimates. “Select HPS” and “Select BRFSS” data are estimates from select states that used the 2022 BRFSS SOGI module (https://www.cdc.gov/brfss/questionnaires/modules/category2022.htm), excluding Guam. “U.S. HPS” and “U.S. NHIS” estimates include all 50 states and the District of Columbia. Panel 1 shows the distribution of all categories. Panel 2 includes the 95% confidence intervals on a truncated scale. All estimates are weighted as directed in the respective user manuals (NCCDPHP 2023a; NCHS 2023; U.S. Census Bureau 2022).

Relative to the differences between the U.S. HPS and U.S. NHIS estimates, the Select HPS and Select BRFSS estimates are more closely aligned. Across categories, the HPS generated higher estimates of the nonstraight population (all responses besides straight) than both the BRFSS (13.6% vs. 12.1%) and NHIS (13.9% vs. 7.2%). However, an exception to this was the Select HPS estimate for “something else,” which was slightly lower than that of the Select BRFSS (2.0% vs. 2.2%). There were also more missing responses for Select BRFSS (2.2%) than there were for Select HPS (1.8%).

In Figure 2, we use a similar method to estimate the adult LGBT population using estimates from the HPS and BRFSS. However, we also integrated 2022 estimates released from Gallup collected via phone interviews (Jones 2023). The Gallup question asked, “Which of the following do you consider yourself to be?” It allowed respondents to select as many options as apply: “(1) straight or heterosexual, (2) lesbian, (3) gay, (4) bisexual, and (5) “transgender.” The 95% confidence intervals are displayed for each estimate. For the Select BRFSS, the estimated LGBT adult population was 6.7%; for the Select HPS, it was 8.4%. U.S Gallup estimates pointed toward an estimated 7.2%, while the U.S. HPS indicated an estimated 8.5%.

Fig. 2.

2022 adult LGBT estimates. “Select HPS” and “Select BRFSS” data are estimates from select states that used the 2022 BRFSS SOGI module (https://www.cdc.gov/brfss/questionnaires/modules/category2022.htm), excluding Guam. “U.S. HPS” and “U.S. Gallup” estimates include all 50 states and the District of Columbia. Gallup estimates are from Jones (2023). The 95% confidence intervals of the estimates are also provided. HPS and BRFFS are weighted as directed in the respective user manuals (NCCDPHP 2023a; U.S. Census Bureau 2022).

Our subsequent analyses of SOGI responses according to birth cohort using the HPS should be considered alongside these triangulated estimates. This research note should not be seen as providing definitive SOGI estimates by cohort, rather our findings should be viewed as an initial step that calls for further exploration as more data sources with appropriate measurement strategies become available within the U.S. population-level data infrastructure, acknowledging that national surveys often produce differing demographic estimates, as in the case of estimating the prevalence of cohabitation (Manning et al. 2019).

Data and Methods

The Household Pulse Survey is an experimental data product initiated during the COVID-19 pandemic to measure how the pandemic affected U.S. households (U.S. Census Bureau 2022). For the sampling strategy, the HPS consists of phases that include weekly time spans and is designed to rapidly release estimates (U.S. Census Bureau 2022). The HPS produces estimates at three geographic levels: the 15 largest Metropolitan Statistical Areas (MSAs), the 50 states and Washington D.C., and the national level (U.S. Census Bureau 2022). In July 2021, the HPS started to ask respondents about their sexual and gender identities (Anderson et al. 2021).

To generate a sample surveyed within the same year for the triangulation of estimates as outlined earlier, we targeted all data collected in 2022. We pooled data collected in Week 42 (between January 26^th and February 7^th, 2022) through Week 52 (between December 9^th and December 19^th). The pooled sample encompassed 10 weeks and included 678,222 respondents. The final sample was restricted to those born between 1945 and 2004 (n = 645,720 respondents) to construct 10-year birth cohorts. Analyses were weighted with replicate weights according to Census Bureau weighting specifications, with appropriate adjustments for pooling multiple waves (U.S. Census Bureau 2022).

For sexual identity, we examined the direct responses to the measure. Respondents were asked, “Which of the following best represents how you think of yourself?” Available responses were: “(1) gay or lesbian, (2) straight, that is, not gay or lesbian, (3) bisexual, (4) something else, and (5) I don’t know.” Because respondents were able to skip the question, we included a sixth option of (6) missing.

Gender identity is measured in the HPS using a two-step procedure. Respondents were first asked, “What sex were you assigned at birth, on your original birth certificate?” Response options were “(1) male or (2) female.” Respondents who did not answer the question were allocated a sex value using a hot-deck procedure and flagged in a separate variable. For their gender, respondents were asked, “Do you currently describe yourself as male, female, or transgender?” Response options were “(1) male, (2) female, (3) transgender, and (4) none of these.” As for sex assigned at birth, respondents could skip the question. However, unlike sex assigned at birth, respondents who were missing on their gender identity were not allocated a value by the Census Bureau. After answering both their sex assigned at birth and gender identity, respondents whose answers differed on the sex and gender measures were asked a confirmatory question on whether the responses were accurate. If the respondents answered no, they were asked the questions again.

We constructed an indicator that signaled whether respondents’ sex assigned at birth differed from their gender identity. For respondents who reported a sex assigned at birth that was the same as their current gender identity, we labeled them as either a cisgender man or a cisgender woman. Respondents who directly selected transgender on the gender measure were categorized as such. Assumptions regarding the gender identity of the respondents who selected transgender should be avoided, as the measurement strategy used in the HPS cannot disaggregate between respondents who for instance identify as trans men or trans women and those who identify as trans and nonbinary.

The utility of employing a two-step measure for gender identity is the ability to detect respondents who may have a sex assigned at birth that is different from their current gender identity but who may not select transgender and instead may select male or female. Given that the HPS does not employ a select-all option for the gender identity question, it cannot be assumed that respondents who selected a gender different from their sex assigned at birth rather than selecting transgender do not identify as transgender, as the force-choice response restricts such conclusions. For respondents who had a sex at birth that was different from their current gender identity and selected male or female and not transgender, we categorized them as DSABGI (different sex assigned at birth and gender identity).

Imputed sex values can lead to overestimates of the gender-diverse population, as nearly half of those who have a sex assigned at birth different from their reported current gender identity and did not select transgender have an allocated sex value. The allocated sex value procedure was independent of the specified value on gender identity, with differing responses between the two measures not going through a verification process. Thus, caution is warranted when including responses with an allocated sex value when studying gender-diverse populations using the two-step procedure. We constructed two additional indicators, one for those who had a sex allocated and a second for those who had a missing gender response. For those who had a missing gender response and an allocated sex, we prioritized the allocated sex. Thus, we constructed a seven-category indicator for responses: (1) cisgender woman, (2) cisgender man, (3) transgender, (4) none of the available gender identities, (5) DSABGI, (6) allocated sex, and (7) missing gender response.

Birth cohort was based on the respondent’s birth year and coded into six 10-year birth cohorts: 1995–2004, 1985–1994, 1975–1984, 1965–1974, 1955–1964, and 1945–1954. Significant differences across cohorts were tested with the youngest cohort (1995–2004) as the reference category.

Results

Table 2 presents responses to the sexual identity question across birth cohorts, with 95% confidence intervals in brackets. Significantly greater shares of younger birth cohorts reported nonstraight responses, including “gay/lesbian,” “bisexual,” “something else,” and “don’t know.” For instance, in the 1945–1954 birth cohort, 93.23% reported a straight sexual identity, compared with 68.87% of the youngest cohort. The increase in bisexual responses contributed the most to the growth in nonstraight responses, from 0.81% in the oldest cohort to 15.93% in the youngest. Older cohorts were more often missing on the sexual identity question compared with younger cohorts. The share of respondents who did not fit into the specific labels offered grew significantly—4.87% of the youngest cohort responded with “something else” and 4.15% responded with a “don’t know” response, compared with only 0.71% and 1.03%, respectively, of the oldest cohort.

Table 2.

Weighted percentages of sexual identity responses across six birth cohorts (1945 to 2004), with 95% confidence intervals

Sexual Identity	Birth Cohort
	1995–2004	1985–1994	1975–1984	1965–1974	1955–1964	1945–1954
Straight	68.87	80.81***	87.95***	90.84***	91.76***	93.23***
	[68.16, 69.58]	[80.31, 81.30]	[87.66, 88.24]	[90.42, 91.24]	[91.46, 92.04]	[92.89, 93.57]
Gay/Lesbian	5.42	4.64***	2.98***	2.75***	2.28***	1.46***
	[5.05, 5.82]	[4.38, 4.92]	[2.81, 3.16]	[2.58, 2.92]	[2.12, 2.45]	[1.37, 1.57]
Bisexual	15.93	7.96***	3.65***	1.74***	1.16***	0.81***
	[15.38, 16.50]	[7.68, 8.26]	[3.49, 3.82]	[1.61, 1.88]	[1.02, 1.32]	[0.71, 0.93]
Something Else	4.87	3.12***	1.82***	1.21***	0.88***	0.71***
	[4.51, 5.27]	[2.90, 3.35]	[1.66, 2.00]	[1.08, 1.37]	[0.79, 0.98]	[0.59, 0.85]
Don’t Know	4.15	2.39***	2.03***	1.57***	1.33***	1.03***
	[3.79, 4.55]	[2.16, 2.64]	[1.88, 2.19]	[1.41, 1.74]	[1.12, 1.48]	[0.88, 1.20]
Missing	0.75	1.08**	1.57***	1.90***	2.60***	2.75***
	[0.57, 0.97]	[0.96, 1.22]	[1.43, 1.73]	[1.70, 2.11]	[2.39, 2.82]	[2.49, 3.03]
n	38,985	103,423	125,638	120,395	142,721	114,558

Notes: Significance testing was completed by comparing each cohort with the 1995–2004 cohort. Confidence intervals are shown in brackets. Analyses were weighted using Census Bureau–designed replicate weights (U.S. Census Bureau, 2022).

Source: Household Pulse Survey, 2022.

^**p < .01

^***p < .001

Table 3 provides the gender identity responses across the six birth cohorts, along with 95% confidence intervals. The shares of transgender identification, DSABGI, and “none of these” responses were significantly higher in the youngest cohort compared with the oldest cohort. For instance, 2.49% of the 1995–2004 birth cohort identified as transgender compared with 0.12% of the 1945–1954 cohort. Individuals in the youngest cohort more often replied as not belonging to any of the offered gender identities relative to those in older cohorts. For example, 3.44% of the 1995–2004 cohort selected “none of these,” as did 0.67% of the 1945–1954 cohort. Furthermore, the share with a missing gender response, as well as those in the DSABGI with an allocated sex category, was generally significantly higher in the older cohorts. The exception to this was that there was no difference between the 1995–2004 and 1985–1994 cohorts for the DSABGI with an allocated sex category.

Table 3.

Weighted percentages of gender identity responses across six birth cohorts (1945 to 2004), with 95% confidence intervals

Gender Identity	Birth Cohort
	1995–2004	1985–1994	1975–1984	1965–1974	1955–1964	1945–1954
Cis Man	48.57	45.87***	47.36**	48.02	46.76***	44.36***
	[47.95, 49.20]	[45.43, 46.30]	[46.92, 47.81]	[47.61, 48.43]	[46.36, 47.15]	[44.00, 44.73]
Cis Woman	44.15	50.33***	49.83***	49.04***	49.98***	52.67***
	[43.47, 44.82]	[49.91, 50.75]	[49.37, 50.28]	[48.68, 49.41]	[49.58, 50.39]	[52.26, 53.07]
Transgender Identity	2.49	1.01***	0.29***	0.28***	0.13***	0.12***
	[2.27, 2.73]	[0.89, 1.14]	[0.23, 0.35]	[0.21, 0.37]	[0.09, 0.20]	[0.07, 0.19]
DSABGI	0.77	0.26***	0.12***	0.06***	0.06***	0.05***
	[0.64, 0.91]	[0.21, 0.33]	[0.08, 0.17]	[0.04, 0.10]	[0.05, 0.09]	[0.03, 0.10]
DSABGI Allocated	0.03	0.07	0.10*	0.14**	0.18***	0.23***
	[0.01, 0.08]	[0.04, 0.11]	[0.07, 0.15]	[0.10, 0.19]	[0.14, 0.23]	[0.18, 0.30]
None of These	3.44	1.67***	1.30***	1.10***	0.95***	0.67***
	[3.16, 3.75]	[1.52, 1.84]	[1.16, 1.47]	[0.99, 1.24]	[0.84, 1.06]	[0.58, 0.78]
Missing	0.55	0.80*	1.03***	1.36***	1.94***	1.90***
	[0.43, 0.70]	[0.69, 0.92]	[0.89, 1.13]	[1.20, 1.54]	[1.78, 2.12]	[1.73, 2.08]
n	38,985	103,423	125,638	120,395	142,721	114,558

Notes: Significance testing was completed by comparing each cohort with the 1995–2004 cohort. Confidence intervals are shown in brackets. Cis = cisgender. DSABGI = different sex assigned at birth and gender identity, which identifies respondents who selected male or female for their gender identity but have a different sex assigned at birth and chose not to identify as transgender. DSABGI allocated individuals are those who had a sex response imputed by the Census Bureau. Analyses were weighted using Census Bureau–designed replicate weights (U.S. Census Bureau, 2022).

Source: Household Pulse Survey, 2022.

^*p <. 05

^**p < .01

^***p < .001.

Discussion

Over recent decades, there has been a significant expansion of the population-level data structure supporting research on LGBTQ+ populations (Baumle 2018; Black et al. 2000; Gates 2013; Gates and Ost 2004; Julian, et al. 2024). As the integration of SOGI questions into population-level data continues, this research note highlights the need for continued innovation in the measurement approaches to reflect the changing ways in which individuals self-identify. Our findings underscore that a higher proportion of younger birth cohorts are more likely to select “something else” for their sexual identity and “none of these” for their gender identity. We acknowledge that there are multiple rationales for selecting these responses, hindering the interpretation of the cohort differences observed in the findings. However, we argue that the increased prevalence of these responses likely reflects the growth in the identification of identities outside of those explicitly captured on surveys among younger sexual- and gender-diverse individuals. This interpretation is consistent with other work that has examined the changing ways in which younger cohorts self-identify (e.g., Brown 2022; Goldberg et al. 2020; Jones 2023; Lagos 2022; Morandini et al. 2017;Watson et al. 2020). Because direct methods to examine the meaning of these responses are limited, further inquiry is needed as more data become available. Still, these findings underscore the importance of researchers knowing the meaning behind “something else” and “none of these” responses as they grow in frequency. To this end, adopting the recommendations put forth by the NASEM committee would ameliorate the current limitations surrounding the “something else” category.

Nonetheless, there are at least three limitations to the NASEM recommendations that warrant further investigation as SOGI measurement practices evolve. First, as pointed out in other research, the use of single-choice responses in measures assumes that respondents identify with a singular identity (Suen et al. 2020; Tabler et al. 2022). More investigation is needed regarding a select-all option that applies to categorization in SOGI questions. Second, the selection of “don’t know” responses, like “something else” responses, may have multiple rationales that make interpretation difficult. “Don’t know” responses may be indicative of respondents not understanding the question, but may also be chosen by those who are questioning their identity. The NASEM recommendations do not include an explicit “questioning” response, so it remains unclear if those who are questioning their identity would opt for “I use a different term” and provide a text response of “questioning” rather than selecting “don’t know”; moreover, this issue may be particularly pervasive among younger populations. Third, the NASEM report does not recommend the inclusion of additional sexual- and gender-diverse identities outside of LGBT and instead suggests text responses to capture all nonlisted identities. Although it is important that respondents share a common understanding of the provided categories to reduce error, it is also vital to acknowledge that reliance on text responses alone may be prone to biases in decoding and interpretation, and these text responses may not be available to all researchers. Thus, we believe text boxes may become a less sustainable strategy as the use of the category grows. Therefore, consideration of other practices must continue to be explored and tested. Despite these limitations, the NASEM recommendations provide an initial framework for the innovation of the SOGI data infrastructure given the cohort differences in observed response patterns.

This research note demonstrates the importance of continuing to innovate methods for measuring sexual and gender identity given the evolving nature of self-identification across birth cohorts. We highlight that the traditional measures used to identify LGBTQ+ individuals may work for older cohorts of the sexual- and gender-diverse population but likely result in the exclusion of larger shares of younger cohorts, thus erasing their experiences. By developing more precise SOGI measurement, we can generate more accurate demographic estimates of the LGBTQ+ population, ensure that younger cohorts are not excluded from analysis of LGBTQ+ health and well-being, conduct more comprehensive investigations of health and well-being among subgroups, and gain a deeper understanding of a growing population in the United States. Moreover, establishing nuanced measurement is linked to understanding how policies targeted at addressing inequalities are operating in the contemporary context and informing the enforcement of civil rights and employment opportunities. Until a full census of the U.S. population includes SOGI questions, the estimates presented here represent part of an initial effort to examine cohort differences in responses. The continued expansion of SOGI questions into population-level federal surveys will further enable triangulation of estimates and findings across surveys, informing both research and policy alike. As the demographic landscape shifts, it is essential to ensure that the population-level data infrastructure keeps pace.