Relying on an “Other” Category Leads to Significant Misclassification of Sexual Minority Participants

Abstract

Purpose: This study assessed which sexual minority participants selected “Other” if their identity was absent.

Methods: This was an online survey; 905 participants saw a limited set of sexual orientation options (bisexual, gay/lesbian, heterosexual, other), and later in the same survey, saw an expanded list.

Results: Twenty-one percent of participants chose different orientation labels across questions. When not presented with a “mostly heterosexual” option, 78% of mostly heterosexual participants chose “heterosexual”; 3% chose “other.” However, when not presented with an “asexual” label, 100% of asexual participants chose “other.”

Conclusion: These findings suggest that “other” categories could misclassify a substantial proportion of sexual minority participants.

Introduction

Many current surveys include some measure of sexual orientation in their demographics. Although most researchers acknowledge that there are increasing numbers of young people who identify as something other than heterosexual, bisexual, or lesbian/gay (e.g., mostly heterosexual, queer),^1–3 there are also concerns regarding respondent fatigue if participants are offered demographics with too many choices.^4,5 Using an “Other” category to capture additional identities could address this, but “Other” labels can be stigmatizing.^3,6,7 Moreover, offering an “Other” option is only helpful if people choose this label when their preferred identity is not present. However, there is evidence that some sexual minority individuals, particularly “mostly heterosexual” people, may instead choose the closest label available, potentially contributing to misleading findings.

One study with a sample of 2077 American college students asked participants to classify themselves using both a 3-category question (heterosexual, lesbian/gay, and bisexual) and a 5-category question (only lesbian/gay, mostly lesbian/gay, bisexual, mostly heterosexual, only heterosexual). This study found that ∼9% of the total sample changed their answers between questions, with the majority of these changes being from “Heterosexual” to “Mostly heterosexual.”⁸ This study suggests that mostly heterosexual participants may simply choose from available options. However, as this study did not include an “Other” option, the question remains: is the “Other” option a viable way to identify sexual minority participants?

In another study using the National Health Interview Survey, participants were presented with a question about their sexual orientation that included a limited set of responses (lesbian/gay, straight, bisexual, something else, or I don't know).⁹ Participants who selected “something else” were then asked what was meant by “something else,” including options such as “I do not use labels to identify myself,” and “I mean something else.” This study found that a higher proportion of cisgender men selected “something else” in both questions, suggesting that some gender/sexuality groups may be more likely to be missed by an “Other” category. However, this study only followed up with participants who initially selected “something else” and did not examine which participants might settle for a listed option even if it did not fully capture their identity.

In this study, I examined congruence in sexual orientation labels when participants in a typical psychology participant pool were initially offered a limited list with an “Other” option, and then later offered a broader array of categories with more specificity. I also explored possible predictors of noncongruence, that is, what predicts changing one's orientation label within the same survey.

Methods

Procedures

Two questions on sexual orientation were embedded in an online prescreening survey offered to all participants in the University of Nebraska-Lincoln (UNL) psychology participant pool (Table 1). The participant pool included a sample of undergraduates taking courses in the psychology department who can receive course credit for participating in research. To match participants with possible research opportunities, the department administers a prescreening survey to identify demographics and eligibility for different studies. The prescreen survey included a wide range of topics (see Supplementary Appendix SA1 for details). This study includes data from two rounds of the survey conducted in Fall 2019 and Spring 2020, with identical sexual orientation questions in both rounds.

Table 1.

Simple and Expanded Questions on Sexual Orientation, as They Were Presented to Participants

Question	Wording	Choices in the order presented
Simple question (presented first)	Please indicate your sexual orientation	Bisexual Gay/lesbian Heterosexual Other
Expanded question (presented second)	Thinking about all of your sexual attractions, sexual fantasies, and lifetime history of sexual partner(s), please select which of these terms best describes you	Exclusively heterosexual Mostly heterosexual Bisexual Mostly gay/lesbian Exclusively gay/lesbian Asexual Other (please describe) Choose not to respond

Measures

Participants were asked a sexual orientation question at the beginning of the survey, which presented the following options: bisexual, gay/lesbian, heterosexual, and other. At the end of the same survey (∼30 minutes later), participants were presented with an expanded orientation question that included the following options: exclusively heterosexual, mostly heterosexual, bisexual, mostly gay/lesbian, exclusively gay/lesbian, asexual, other, and choose not to respond. In both instances, participants were asked to select a sexual orientation; however, in the expanded question, this was clarified to include the participants' patterns of sexual attractions, fantasies, and lifetime history of sexual partners. I coded responses as matching if participants indicated the same response both times (e.g., “Heterosexual” in the first question matching to “Exclusively heterosexual” in the second question).

Participants completed a battery of demographics including age, race/ethnicity, and political affiliation (Table 2). Gender was assessed with the question “what is your sex/gender,” and participants could choose from the following: Female/Woman, Male/Man, Transgender M-F, and Transgender F-M. Participants were also asked their level of sexual desire in the past month using the sexual desire subscale of the Female Sexual Function Index (FSFI). The FSFI is brief self-report measure of sexual function that has been validated in men and women, and in both heterosexual and sexual minority participants.^10,11 Scores on the desire subscale range from 1 (little to no desire) to 5 (very high desire). A measure of sexual desire was included to test the myth that participants who identify as mostly heterosexual or bisexual are simply heterosexuals with very high sexual interest.^12,13

Table 2.

Demographics of Study Participants

	Consistent orientation label (n = 711)		Changed orientation label (n = 194)		Total (N = 905)		χ2	p
	n	%	n	%	n	%
Sex/gender
Cisgender woman	509	72	163	84	672	74	30.29	<0.001
Cisgender man	201	28	27	14	228	25
Transgender woman or man	0	0	4	2	4	0
Data missing	1	0	0	0	1	0
Relationship status							0.09	0.763
Single	398	56	110	57	508	56
In a relationship	308	43	81	42	389	43
Data missing	5	1	3	2	8	1
Race/ethnicity							4.83	0.438
White non-Hispanic	543	76	137	71	680	75
Black	26	4	9	5	35	4
Hispanic	45	6	20	10	65	7
Asian/Pacific Islander	78	11	24	12	102	11
Native American	4	1	1	1	5	1
Other	14	2	3	2	17	2
Data missing	1	0	0	0	1	0
Political affiliation							41.87	<0.001
Democrat	219	31	97	50	316	35
Republican	223	31	22	11	245	27
No affiliation	238	33	60	31	298	33
Other	30	4	14	7	44	5
Data missing	1	0	1	1	2	0
Religion							33.5	<0.001
Catholic	222	31	37	19	259	29
Protestant	159	22	36	19	195	22
Atheist	54	8	23	12	77	9
Agnostic/believe in God	165	23	79	41	244	27
Other	109	15	18	9	127	14
Data missing	2	0	1	1	3	0
Parents' highest education							26.52	<0.001
Less than high school	18	3	20	10	38	4
High school graduate	56	8	14	7	70	8
Some college	101	14	22	11	123	14
College graduate	271	38	74	38	345	38
Some graduate/professional school	49	7	6	3	55	6
Graduate/professional school graduate	213	30	58	30	271	30
Native language							2.88	0.411
English	615	86	161	83	776	86
Chinese	33	5	12	6	45	5
Spanish	19	3	9	5	28	3
Other	38	5	11	6	49	5
Data missing	6	1	1	1	7	1
Employment status							0.97	0.615
Student full time	485	68	127	65	612	68
Employed part time	211	30	61	31	272	30
Employed full time	12	2	5	3	17	2
Data missing	3	0	1	1	4	0
	M	SD	M	SD	M	SD	t	p
Age	19.4	1.86	20.1	2.88	19.6	2.14	2.88	0.004
Sexual desire score	3.55	1.05	3.43	1.22	3.52	1.09	−1.15	0.252

The sexual desire score was derived from the Female Sexual Function Index. Scores on the subscale can range from 1 (little to no desire) to 5 (very high desire).

M, mean; SD, standard deviation.

Study participants

A total of 913 students completed the survey; of these, 905 provided data for both orientation questions. Participants were predominantly White (75%) cisgender women (n = 74%, Table 2). Participants provided informed consent, and all procedures were approved by the UNL Institutional Review Board.

Statistical analyses

To explore possible predictors of participants changing sexual orientation responses, I selected those factors most likely to be present in typical demographics (e.g., age, sex/gender). I first assessed each variable separately in chi-squared analyses of independence (for categorical variables) or t-tests (for continuous variables); then, I included all significant predictors in a logistic regression to assess if they remained significant when controlling for shared variance. These analyses were exploratory, and intended to identify subgroups for whom the use of an “Other” category might be particularly problematic, which may inform intersectional approaches to sexual identity categorization across diverse populations. Deidentified data supporting all analyses are available on the Open Science Framework.¹⁴ All analyses were conducted in R version 4.0.¹⁵

Results

Demographics and changes in orientation labels

Twenty-one percent of participants (n = 194) changed their sexual orientation choice across questions (Table 3). Of these changes, the majority were from “Heterosexual” to “Mostly heterosexual” (60% of total changes), followed by “Bisexual” to “Mostly heterosexual” (14% of total changes). Of the participants who chose “Mostly heterosexual” in the second question, only 3% had initially indicated “Other.” The majority (78%) of mostly heterosexual participants had selected “Heterosexual,” with 19% selecting “Bisexual.” In contrast, of the participants who indicated “Asexual” in the second question, 100% had initially indicated “Other.”

Table 3.

Responses to the Sexual Orientation Questions, Before and After Clarification

	Exclusively heterosexual		Mostly heterosexual		Bisexual		Mostly gay/lesbian		Exclusively gay/lesbian		Asexual		Other		Missing		Choose not to respond
	n	%	n	%	n	%	n	%	n	%	n	%	n	%	N	%	n	%
Heterosexual	667	99	117	78	3	8	0	0	0	0	0	0	1	17	2	25	9	82
Bisexual	0	0	28	19	34	85	8	62	1	11	0	0	2	33	1	13	0	0
Gay/lesbian	1	<1	0	0	0	0	3	23	8	89	0	0	1	17	0	0	0	0
Other	1	<1	5	3	3	8	2	15	0	0	7	100	2	33	0	0	2	18
Missing data	0	0	0	0	0	0	0	0	0	0	0	0	0	0	5	63	0	0

Rows indicate the initial response; columns indicate the second response. Cells in which participants responded consistently across both questions are listed in bold. Percentages are calculated within column.

Predictors of changing sexual orientation response

Changing one's response to the sexual orientation question was significantly predicted by being a cisgender woman, transgender man or woman; religious identity as agnostic or “belief in God without specific affiliation”; Democratic political affiliation; older age; and lower parental education (Table 2). The logistic regression model that included all these predictors was significant [Model χ²(16) = −385.06, p < 0.001], and explained 14% of the total variance (McFadden's Pseudo R² = 0.139). All entered predictors (age, sex/gender, religion, political affiliation, and parent's highest education) remained significant in the same directions as the original separated analyses (Table 4). Of note, sexual desire was not a significant predictor of changing to the mostly heterosexual or bisexual label.

Table 4.

Predictors of Changing One's Orientation Label

Predictor	B	SE (B)	z	p
Intercept	−1.41	0.85	−1.65	0.10
Age	0.13	0.04	3.26	<0.001
Sex/gender
Cisgender woman (referent)
Cisgender man	−1.03	0.25	−4.04	<0.001
Transgender man or woman	15.53	413.81	0.04	0.97
Religion
Agnostic/no specific affiliation (referent)
Catholic	−1.01	0.25	−4.03	<0.001
Protestant	−0.35	0.25	−1.38	0.17
Atheist	−0.17	0.32	−0.54	0.59
Other	−1.08	0.32	−3.39	<0.001
Political affiliation
Democrat (referent)
Republican	−1.28	0.28	−4.61	<0.001
No affiliation	−0.42	0.21	−2.01	0.04
Other	0.05	0.40	0.13	0.90
Parents' highest education
Less than high school (referent)
High school graduate	−1.53	0.50	−3.07	<0.001
Some college	−1.75	0.46	−3.79	<0.001
Graduated college	−1.31	0.40	−3.25	<0.001
Some graduate or professional school	−2.13	0.59	−3.62	<0.001
Finished graduate or professional school	−1.30	0.41	−3.15	<0.001

B, unstandardized regression weight; SE (B), standard error of the unstandardized regression weight.

Discussion

The purpose of this study was to (a) examine the degree to which the “Other” option successfully captured sexual minority participants whose identity label was not presented, and (b) explore variance in answers when the same participants were presented with a simple sexual orientation question versus an expanded list of options. I found that more than a fifth of participants changed their orientation labels when given more options, with the majority of these being a change to “Mostly heterosexual.” However, only a very small proportion of people who later identified as “Mostly heterosexual” initially chose the “Other” option, suggesting that reliance on an “Other” category substantially underestimates the degree of diversity of sexual orientation in typical college samples.

The fact that ∼17% of participants identified as “mostly heterosexual” highlights the importance of including this label in surveys of young adults' sexual orientation. Mostly heterosexual individuals are the largest sexual minority group¹⁶ but may not identify themselves unless specifically cued to do so. Indeed, the data suggest that mostly heterosexual individuals will generally identify as heterosexual if that is the closest option provided, which could add significant error variance to analyses using those data.

In contrast to the mostly heterosexual participants, all participants who identified as asexual chose the “Other” label when not provided with the “asexual” choice. These findings need to be replicated, as the asexual group was quite small (less than 1% of the total sample). However, tentatively, these findings could point to significant differences in how asexual and mostly heterosexual participants perceive entitativity of their orientation—that is, the degree of cohesiveness of their orientation group. It is possible that mostly heterosexual individuals perceive their identity as sharing boundaries with heterosexual or bisexual labels (i.e., low entitativity), whereas asexual individuals perceive their identity as qualitatively distinct (i.e., high entitativity). Some research suggests that the degree of perceived entitativity of a sexual minority group label predicts the degree of identity satisfaction in people who identify with that label.¹⁷ Future work should investigate the qualities of asexual and mostly heterosexual identities that may place them at opposite ends of the entitativity spectrum.

Of interest, 13 participants who changed their responses selected an option that had previously been presented (e.g., changed from initial “Heterosexual” to later “Bisexual”). One possibility is that participants changed in error (as was likely the case for the one individual who initially indicated “gay/lesbian” and then later “exclusively heterosexual”). However, another possibility is that by asking the same question twice, participants responded to perceived experimenter demands: that is, in presenting an expanded view of the construct of sexual orientation, participants may have perceived that the researchers had a broadly inclusive view of sexual orientation, which in turn encouraged them to respond in ways that align with that perceived view. This interpretation is supported by the fact that changing one's answer was more likely among people with other identities typically associated with liberal views (e.g., agnostic, Democrat). However, it is equally likely that liberalness is associated with belief in sexual orientation as a spectrum^18,19 and higher likelihood of identifying with a mostly heterosexual orientation.¹⁹ It is also possible that by specifically cuing participants into the various dimensions of sexual orientation (e.g., attractions, fantasies, and behaviors), participants may have considered a wider range of factors that led them to reconsider their identity label.

Limitations

These findings should be interpreted with the study's limitations in mind. All the participants were in college, and the sample was predominantly White non-Hispanic, female, and cisgender; thus, generalizability to a more diverse or noncollege sample is unknown. In addition, because these data were derived from a mass screening survey used by many research groups, I was unable to alter certain aspects of how the demographics were presented. Specifically, the limited choice question was always presented first, and the expanded question second. It is possible that mostly heterosexual individuals may have been more likely to choose the “Other” option in the limited choice question if they had been primed by seeing this as an option earlier. That said, these data are still relevant for understanding the identity labels chosen by mostly heterosexual participants in typical demographics without extensive cuing. Another important limitation was that even the expanded question did not include options for identities that are emerging as particularly relevant for young adult samples, such as pansexual or queer. Finally, an interesting avenue for future research will be qualitative interviews investigating the reasons why participants changed their responses and their interpretation of the different items.

Conclusion

I found that 21% of participants gave different answers when presented with a very basic question about their sexual orientation label versus an expanded question that included attraction and behavior. Most of these changes were to select a “mostly heterosexual” label. However, in the absence of a “mostly heterosexual” label, the majority of mostly heterosexual participants chose another label rather than the “Other” category. A growing body of evidence suggests that the mostly heterosexual orientation is the largest sexual minority population,¹⁶ and as such, should be considered as important as other commonly assessed labels such as “bisexual.” Moreover, an increasing body of research suggests that, relative to heterosexual people, mostly heterosexual people experience significantly higher rates of mental and physical health problems,²⁰ are at higher risk for experiencing sexual trauma,^21,22 and have underserved sexual health needs.¹⁰ The current findings suggest that unless we specifically target mostly heterosexual identities in our measures, we will continue to miss important effects that could lead to better interventions for these significant public health concerns. The very moderate increase in length and complexity of one's survey by including an expanded orientation question stands to reduce mischaracterizing more than one-fifth of one's sample, and could contribute significantly to research aiming to identify health disparities in sexual minority young adults.

Author's Contributions

T.K.L. was the sole author, and was responsible for all components of the article (research design and implementation, data analysis, and article preparation and editing).

Disclaimers

The content is solely the responsibility of the author and does not necessarily represent the official views of the National Institutes of Health or the University of Nebraska.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This research was supported by internal funds from the UNL Department of Psychology, the Nebraska Tobacco Settlement Biomedical Research Development Fund, and the UNL Office of Research and Economic Development. This work was also supported by the National Institute of General Medical Sciences of the National Institutes of Health (P20GM130461) and the Rural Drug Addiction Research Center at the University of Nebraska-Lincoln.

Supplementary Material

Supplementary Appendix SA1