Over the past decade in the United States there has been a push for more inclusive demographic categories, especially for gender identity and sexual orientation, in a variety of contexts, including in medical records (e.g., Cahill and Makadon 2013) and university admission (see Jaschik 2011). However, many proposals for adding additional demographic questions regarding sexual orientation and gender identity have been rejected by national medical associations (Cahill and Makadon 2013) and universities (Jaschik 2015; Stainburn 2013; Windmeyer 2012). There are a variety of different formats presently used to measure gender identity, each with different benefits and pitfalls. It is important for scientists, universities, physicians, businesses, and other institutions to understand how individuals completing such questionnaires feel toward different measures of gender identity and what the source of these opinions are.

In the present research, we investigated preferences for binary versus more expansive gender identity questionnaires among cisgender (participants whose gender identity matches the label assigned at birth) heterosexual; cisgender lesbian, gay, and bisexual (LGB); transgender (participants whose gender identity does not match the label assigned at birth; Tate et al. 2014); and gender non-binary (participants whose gender identity does not conform to the female/male gender dichotomy) samples. We examined what individual differences predicted these preferences and how the use of different gender-identity questionnaire formats would affect attitudes toward institutions.

When discussing the issues of identifying sex and gender in research and applied settings, it is necessary to delineate the difference between sex and gender, especially regarding why researchers need to collect data on either variable (i.e., what purpose in the research do either sex or gender have). Sex is generally conferred by medical personnel at birth and commonly refers to the genital anatomy of an individual (Tate et al. 2013), whereas gender refers to an individual’s psychological sense of self and their expected behavioral gender roles within their sociocultural context (Bussey and Bandura 1999). There are many situations in which individuals are asked about their sex and/or gender, and many areas of social research typically require examining sex or gender differences (Tate et al. 2013). U.S. Census data has collected national sex indexes since the 1900 census (United States Census 2017) to track group characteristics, such as employment and potential discriminatory actions (American Community Survey 2016). In U.S. medical research, sex and gender groups are often delineated to assess health risks, longevity and protective health factors (e.g., National Health Interview Survey 2014; Behavioral Risk Factor Surveillance System 2015), and track sexually transmitted infections (Centers for Disease Control and Prevention 2002).

Questions about gender can be asked in many different ways, with two to hundreds of possible gender-identity categories from which individuals can choose. Proponents of implementing expanded demographic categories for sexual orientation and gender identity argue that the identification of transgender and gender non-binary individuals in population data is important for public policy, sciences, and medical treatment (Deutsch et al. 2013; Moscoe 2014; Reisner et al. 2014; Ridolfo et al. 2011; Sell and Becker 2001; Sexual Minority Assessment Research Team 2009; Wylie et al. 2010). The identification of transgender and gender non-binary individuals allows for data collection across a variety of research areas including physical health, mental health, and personal experiences, such as those of marginalization and discrimination (Dickey et al. 2016; Reisner et al. 2014; Wylie et al. 2010). Proponents argue that without the ability to self-identify in questionnaires, surveys, and other data collection procedures (i.e., when only presented with binary gender identity options), the voices of transgender and gender non-binary individuals tend to be excluded or ignored (Dickey et al. 2016).

Additionally, universities that have implemented expanded demographic categories in admissions surveys purport that identifying LGB and non-cisgender students allows for the creation of special provisions, programs, and communities for LGBT students, as well as including non-cisgender students as an identified and measured group in university and educational climate evaluations (Ivory 2005). Also, allowing students to report their sexual orientation and gender identity sends the message that LGBT students are welcome at that institution (Conron et al. 2008; Windmeyer 2012). Indeed, LGBT students tend to perceive their campus climate as warmer, more supportive, and more welcoming when they are allowed to disclose their sexual and gender identity (Evans 2001; Evans and Broido 2002; Garber 2002), which is associated with more positive academic and psychological outcomes for those students (Rankin 2004; Schmidt et al. 2010). Furthermore, the traditional dichotomous gender format can be difficult for gender minorities to answer and may lead them to feel ignored (Ridolfo et al. 2011; Sexual Minority Assessment Research Team 2009). Research has indicated that misgendering (having one’s gender identity misclassified or not recognized by others) is associated with increased feelings of stigmatization (leading to greater negative affect), as well as lowered appearance self-esteem and identity strength and coherence among transgender individuals (McLemore 2014).

Conversely, opponents of expanded demographic categories purport that the identification of LGBT individuals threatens their privacy and confidentiality and allows for discrimination and bias due to identification as sexual or gender minorities (Cahill and Makadon 2013; Jaschik 2011; Stainburn 2013; Windmeyer 2012). In college admissions, reporting sexual and gender identity may be uncomfortable for young people who are not yet stable in their identities (see Kobliha 2015; Stainburn 2013). These views may not be without merit; social theories of sexual stigma (Herek 2007; Herek et al. 2015; Sánchez and Vilain 2009) and the minority stress model as applied to sexual and gender minorities (Frost and Meyer 2009; Meyer 1995; Meyer 2003) posit that institutionalized heterosexism (i.e., the cultural norm and expectancy of heterosexual orientation and biological gender conformity) may inhibit the self-identification of gender non-conforming individuals due to fear of exposure.

The sexual stigma model (Herek 2007; Herek et al. 2015) suggests that culturally accepted, enacted stigma (i.e., blatant expressions of prejudice and discrimination toward LGBT individuals) may lead gender non-conforming individuals to alter their behavior in order to avoid the experience of prejudice and discrimination based on their gender identity. This decision might include avoidance of identifying themselves as gender non-conforming, especially if they feel that they can “pass” as their ascribed or biological gender (Sánchez and Vilain 2009). Similarly, the minority stress model posits not only that gender and sexual minorities experience unique stressors due to social stigma, but also that appraisals of their environment as potentially dangerous (physically and psychologically) may lead them to conceal their sexual and gender identities to avoid rejection, persecution, and stigmatization (Frost and Meyer 2009; Meyer 2003).

Little research has been conducted addressing the actual benefits or harms of using traditional gender questions versus more expansive demographic categories. Instead, most of the arguments on both sides of the debate are based in conjecture. A few interview studies have assessed the attitudes of LGBT individuals toward more inclusive sexual orientation and gender identity categories, wherein LGBT participants reported that they felt more included and found the more expansive categories easier to complete (Conron et al. 2008; Conron et al. 2014; Ridolfo et al. 2011; SMART 2009). One study, which piloted a more inclusive measure of gender identity (distinguishing sex from gender) with trans- and cis-gender adolescents, found that most had favorable views toward the more inclusive measure (Conron et al. 2008). A focus group of gender-diverse individuals evaluating a single-item versus two-question gender identity format for use in medical records indicated that although the two-question format was preferable given its ability to accurately identify, participants also had concerns about information privacy and involuntary disclosure (Thompson 2016).

Additionally, several studies have investigated the psychometric quality of more expansive measures of gender identity (e.g., Dargie et al. 2014; Gender Identity in US Surveillance Group 2014; Harrison et al. 2012; Joel et al. 2014; Reisner et al. 2014; Tate et al. 2013). The GenIUSS (2014) study and the Institute of Medicine (IOM 2013) reviews suggest that the best practice for asking about gender identity is to use a two-question (i.e., birth sex and current gender identity) format. This format was shown by Tate et al. (2013) to have superior data quality compared to a single-item non-binary format (i.e., male, female, transgender, other). Another study provided LGBT and heterosexual patients at four U.S. health clinics a demographic question for sexual orientation and a two-part gender identity item (Cahill et al. 2014). These researchers found that the items were easy to understand and that the participants felt that they were able to correctly identify their identities.

However, previous studies have not explicitly asked community members across the United States about their preferences toward different gender questionnaire formats or what predicts these preferences. Such research merits further expansion to assess the general public’s attitudes about these more inclusive and expanded categories of gender identity. It is also important to examine how the use of such expanded categories may affect attitudes toward institutions. In general, it seems to be most common in the United States for cisgender persons and cisgender-normed institutions to make assumptions and, ultimately, decisions about the transgender community, particularly when it comes to the matter of reporting and collecting data about gender identity. We are interested in what individual differences may influence attitudes, and thus opinions, about gender identity measures.

Although the field is relatively new, research on prejudice toward transgender individuals suggests there are a variety of individual difference and personality traits associated with increased anti-transgender prejudice and transphobia including right-wing authoritarianism, religious fundamentalism, gender-binary attitudes, political conservatism, religiosity (for cisgender heterosexual women only), anti-egalitarian attitudes, hostile sexism (Nagoshi et al., 2008), as well as Need for Closure, preference for traditional gender roles, and anti-gay, anti-lesbian, and anti-bisexual (LGB) prejudice (Tebbe and Moradi 2012). Additionally, cisgender heterosexual men tend to have more negative attitudes toward transgender individuals than cisgender heterosexual women do (Antoszewski et al. 2007; Gerhardstein and Anderson 2010; Norton and Herek 2013; Tebbe and Moradi 2012). In an experimental clinical setting, cisgender heterosexual male clinicians were less willing to interact with a fictional trans man client (Carroll et al. 2012). These individual difference variables connected to anti-trans prejudice are also likely to predict attitudes toward expanded gender identity questionnaires and institutions that use expanded questionnaires.

The purpose of the present studies is to examine the attitudes of cisgender, transgender, and gender non-binary samples of adults in the United States toward non-binary demographic categories for gender identity compared to the traditional binary gender categories (i.e., male, female). Because little research on this topic has been published, there are several ways personal values may influence attitudes about the presentation of gender categories. First, distinctiveness threat, which stems from social identity theory (Tajfel and Turner 1979; Turner 1999), suggests that people are motivated to maintain intergroup distinctiveness (e.g., distinct criteria for categorizing gender). People who perceive binary gender as highly distinctive may perceive persons who blur those distinctions as threatening and prefer the expanded gender categories because it could allow for the identification of transgressors. Conversely, they may prefer the traditional dichotomous gender categories because acknowledging other categories blurs the distinct lines between binary genders. Second, preferences may be moderated by belief in gender binarism (i.e., there are only two genders, corresponding with biological sex) and need for closure (Kruglanski and Webster 1996), which is associated with transgender prejudice (Tebbe and Moradi 2012) and preferences for social categorization (Kruglanski et al. 2006; Miller et al. 2009; Webster and Kruglanski 1997).

In three studies, we assessed the attitudes of U.S. cisgender men and women across sexual orientations (Study 1), heterosexual cisgender men and women (Study 2), cisgender LGB men and women (Study 3), and transgender and gender non-binary individuals across sexual orientations (Study 3) regarding different formats of gender questionnaires. In Study 1, participants reported their attitudes toward two different gender questionnaire formats: one with the traditional binary options for gender (i.e., male, female) and another with two separate items for birth sex and gender identity. Studies 2 and 3 included a third gender questionnaire format: a single-choice measure with 58 options for gender identity and gender expressions.

Our primary interest in the attitudes of cisgender individuals was to assess how their beliefs and values influenced their reactions to non-binary questionnaire formats. As previously discussed, cisgender persons are often the ones making decisions about how to ask about gender. Thus, understanding what influences their attitudes toward different gender question formats could provide insights into why they approve or disapprove of implementing new formats. Perhaps more important are the opinions of people who identify as transgender and gender non-binary because they are the individuals who, presumably, are excluded when binary questionnaire formats are used. However, it is possible that non-cisgender (e.g., transgender and gender non-binary) individuals prefer not to identify as non-binary (i.e., prefer to identify as male/female only, not as transgender).

Our main relationships of interest were between preferences for the different gender questionnaire formats, attitudes regarding those formats, and individual difference variables that have been shown to be associated with transgender prejudice (i.e., beliefs in the gender binary, distinctiveness threat, political orientation, religiosity, age, cisgender heterosexual men, and less transgender contact). We predicted that greater gender binarism, more conservative political orientation, greater religiosity, cisgender heterosexual men, and those with less transgender contact would prefer the binary format and have more negative evaluations of the non-binary formats. The relationships between distinctiveness threat and format preference and attitudes is more exploratory due to the fact that, as mentioned previously, those high in gender-based distinctiveness threat may prefer the binary format because acknowledging the existence of more than two binary gender identities is threatening; however, they may prefer the non-binary formats because those formats allow for the detection of people who blur and threaten binary gender boundaries, who can then be avoided.

Study 1: Cisgender Sample

To assess the attitudes of cisgender individuals toward different gender questionnaire formats, we recruited Amazon’s Mechanical Turk (mTurk) workers to view and rate two versions of demographic questions for gender: (a) one with binary options only and (b) another with two separate items for biological sex and gender identity, including options for non-binary gender. We also measured distinctiveness threat and gender binarism. We asked three research questions: (a) Which presentation of gender categories (i.e., binary gender versus expanded gender categories) in demographic surveys do cisgender persons prefer?, (b) What are the reasons they prefer one presentation type over the other (e.g., inclusiveness/exclusiveness)?, and (c) Do distinctiveness threat and gender-binary beliefs affect attitudes toward gender category presentation? We predicted that greater gender binarism, more conservative political orientation, greater religiosity, cisgender heterosexual men, and those with less transgender contact would prefer the binary format and have more negative evaluations of the non-binary formats. We also wanted to see how distinctiveness threat would affect which version of the gender questions participants preferred, but because the present study was exploratory, we did not predict a specific direction.

Method

Participants

Fully 376 of Amazon’s Mechanical Turk (mTurk) workers were recruited; however, 41 were excluded for either failing attention checks or for being located outside the United States. One additional participant was excluded who identified as cisgender male based on the expanded format response, but as cisgender female on the demographics item. Six participants who identified as non-cisgender were also excluded from analyses because the purpose of the current study was to assess cisgender participants’ attitudes, and 59 did not respond to the expanded format items and were excluded because their gender identity could not be determined. Thus, a total of 269 (72%) participants were retained for analyses.

Of the 269 analyzed, 254 (94%) responded to the demographic questions: 166 (65.4%) identified as female and 88 (34.6%) as male, with a mean age of 40.63 years (SD = 13.29, range = 21–76). Of those who reported their ethnicity, 203 (75.5%) identified as White, 20 (7.4%) as Black, 15 (5.6%) as Hispanic, 9 (3.3%) as Asian American, and less than 3% identified as either Native American (n = 2), biracial (n = 4), or other (n = 1). The sample was also relatively educated (167, 65.7%, having an Associate’s degree or higher), more liberal (112, 41.7%, liberal-very liberal; 82, 30.5%, moderate; 59, 22%, conservative-very conservative) and less religious (124, 46.1%, non-religious; 76, 28.2%, moderately-highly religious). Gender identity of participants was determined by their responses to our expanded format (the two-item gender measure with biological sex and gender identity; see description of the expanded format measure in the following gender question formats subsection).

Procedure and Measures

Participants responded to an online survey after first providing their consent. All participants viewed and rated both versions of the gender questionnaire formats as non-interactive images, presented in random order. Following each gender questionnaire format, participants rated the perceived inclusiveness of that format, and then were asked how they would respond to the questionnaire format themselves (i.e., interactively responded to the questionnaire format) and provided their format preference. Participants also completed measures of distinctiveness threat and gender binarism, as well as demographic information (i.e., age, sex/gender identity, ethnicity/race, citizenship status, first language, hometown population, political orientation [single item and multi-item: social issues, economic issues, foreign policy issues], political party affiliation, religious affiliation, religiosity, transgender contact), in that order. We also included measures of Need for Closure (Roets and Van Hiel 2011) and additional format rating items assessing perceptions that the formats blurred gender boundaries and were useful for detecting certain people in order to avoid them. For parsimony, these measures are not reported but are available as an online supplement. Participants were paid $.25 for their participation.

Gender Questionnaire Formats

Two formats for gender questionnaires were presented in random order to participants. The binary option simply asked respondents: “What is your gender? (check one)” and provided two response options: Male or Female. The expanded format asked two questions. The first, “What is your current gender identity” (check all that apply),” included seven response options: (a) Male; (b) Female; (c) Female-to-Male (FTM)/Transgender Male/Trans Man; (d) Male-to-Female (MtF)/Transgender Female/Trans Woman; (e) Genderqueer, neither exclusively male or female; (f) Additional Gender Category/(or Other), please specify ________; or (g) Decline to answer. The second question, “What sex were you assigned at birth on your original birth certificate? (check one)” provided three response options: Male, Female, and Declined to Answer.

Perceived Inclusiveness of the Questionnaire Format

Participants rated each format on 12 items assessing how inclusive and appropriate (i.e., “correct”) they felt it was using a Likert-type scale from 1 (strongly disagree) to 5 (strongly agree). Factor analysis revealed that six inclusiveness items created a single factor for both the ratings of the binary and expanded formats: “I think that this way of asking about gender…” “…will make members of the LGBT community feel more included”; “…will avoid making non-binary gendered people worried about facing discrimination”; “…will make non-binary gendered people more comfortable”; “…is more useful for medical reasons (e.g., getting people the right treatment faster)”; “…makes it easier to identify people’s gender correctly”; and “…is better for non-binary gendered people so they don’t feel left out.” Responses were averaged to create a single Inclusiveness Rating score wherein higher scores indicated a rating of the question format as more inclusive. The alpha for the Inclusiveness Rating factor was .83 for the binary format and .77 for the expanded format.

Gender Binarism

This modified scale selected five items from the nine-item Transphobia Scale (Nagoshi et al. 2008) to assess respondents’ preference for binary gender, with each rated on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). The five selected items were: “A person’s genitalia define what gender they are, e.g., a penis defines a person as being a man, a vagina defines a person as being a woman”; “I believe that a person can never change their gender”; “I think there is something wrong with a person who says that they are neither a man nor a woman”; “I believe that the male/female dichotomy is natural”; and “I am uncomfortable around people who don’t conform to traditional gender roles, e.g., aggressive women or emotional men.” Items were averaged into a single composite measure of gender binarism wherein higher scores indicated stronger beliefs in a gender binary. This full scale has shown good reliability in past research (α = .82, Nagoshi et al. 2008; α = .84, Tebbe and Moradi 2012). In the present study, the five-item scale showed good reliability (α = .85).

Distinctiveness Threat

Two items were modified from the Distinctiveness Threat items developed by Warner et al. (2007) to assess feelings of threat from the blurring of binary gender categories. Two statements were rated on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree): “Asking about more than two genders (i.e., male and female) threatens the differences between people who are born female and people who are born male” and “Asking about more than two genders (i.e., male and female) blurs the boundary between female and male genders.” However, the correlation between the items suggested that each item was tapping a different aspect of distinctiveness threat. The second item is somewhat ambiguous, therefore we only used the first item, which better represents the construct of distinctiveness threat.

Format Preference

Following the format ratings and individual difference variables, participants were asked which of the two formats they preferred. Participants were shown a graphic containing the binary format (labeled “Form A”) and the expanded format (labeled “Form B”) and were asked, “Which way of asking about gender would you prefer to see on a standard form (e.g., in a research study, for school applications, at a doctor’s office)?” for which they could select either Form A (binary format) or Form B (expanded format).

Results

Due to the low reliability of some composite scales and in order to avoid Type II errors due to attenuation of the effects, we computed corrected correlations to adjust for the reliability of measures (Anglim 2009; Borenstein et al. 2009; Osborne and Waters 2002):

$$ \mathrm{Corrected}r=\frac{r_{xy}}{\sqrt{\upalpha_{\mathrm{x}}}\sqrt{\upalpha_{\mathrm{y}}}} $$

Several measures consisted of only a single item (e.g., participant gender, distinctiveness threat). For the single-item measures, we used a conservative estimate of reliability, 1.0 (Borenstein et al. 2009). Corrected correlation significance was computed based on Pearson r critical value tables because the corrected correlations were computed by hand (i.e., not using SPSS). We conducted corrected correlations between relevant participant characteristics (i.e., binary gender, political orientation, age, highest level of religiosity, transgender contact) and questionnaire format ratings (i.e., inclusiveness, format preference; see Table 1).

Table 1 Corrected correlations between questionnaire format ratings and participants’ characteristics, Study 1
Full size table

Frequency of Format Preferences

Overall, preferences for the expanded and binary formats were relatively even: 47.5% (124 participants) reported a preference for the expanded format whereas 52.5% (137 participants) reported a preference for the binary format, and eight participants did not respond to the preference question. Preference was analyzed using a binomial test and the test demonstrated that participants did not show a preference for either format. Neither expanded format preference nor binary format preference was selected significantly more frequently than chance (i.e., 50%), binomial test p = .46.

Correlations: Both Question Formats

Perceiving the binary format as more inclusive was associated with higher belief in gender binarism, higher distinctiveness threat, more conservative political orientation, and higher religiosity, as well as less contact with transgender individuals (see Table 1). Binary format preference was associated with greater gender binarism, greater distinctiveness threat, more conservative political orientation, higher religiosity, and less transgender contact (see Table 1). Perceiving the expanded format as more inclusive was associated with cisgender women participants, more transgender contact, lower belief in gender binarism, lower distinctiveness threat, less conservative political orientation, younger age, and lower religiosity (see Table 1).

Logistic Regression

We also conducted a binomial logistic regression to test the effects of gender binarism, distinctiveness threat, political orientation, binary gender (i.e., female = 1, n = 164; male = 0, n = 86), age, transgender contact (no contact = 1, n = 136; contact = 0, n = 114), and religiosity on the likelihood of preferring the binary format (i.e., expanded format preference coded as 0; binary format preference coded as 1). Collinearity statistics indicated that the variables were not problematically correlated with other predictors (i.e., VIF < 10, tolerance > .10; Abu-Bader 2010).

The logistic regression model was significant, χ2(7) = 135.98, p < .001. The inclusion of predictors in the model benefited the model fit (Nagelkerke R 2 = .56), and the model correctly classified 81.6% of cases. Those with high gender binarism beliefs were nearly four times more likely to prefer the binary format (β = 1.36, SE = .29), Wald(1) = 22.31, p < .001, odds ratio = 3.90, 95% CI [2.22, 6.86]. Additionally, those with transgender contact were about twice as likely to prefer the expanded format (β = .71, SE = .38), Wald(1) = 4.05, p = .04, odds ratio = 2.02, 95% CI [1.01, 4.02]. More conservative participants were one and a half times more likely to prefer the binary format (β = .40, SE = .14), Wald(1) = 8.25, p = .004, odds ratio = 1.49, 95% CI [1.14, 1.97]. Additionally, those higher in distinctiveness threat were nearly one and a half times more likely to prefer the binary format (β = .42, SE = .20), Wald(1) = 4.47, p = .03, odds ratio = 1.51, 95% CI [1.03, 2.22].

Discussion

Based on the results of Study 1, it appears that attitudes about gender questionnaire formats are most strongly influenced by gender binarism beliefs and distinctiveness threat, as well as by participants’ gender and ideology (i.e., political orientation, religiosity). However, we were curious as to whether responses to these gender questionnaire formats, especially the expanded format, would differ if we included another format for non-binary gender. In our study, those high in gender binarism and distinctiveness threat tended to rate the binary format as more inclusive than the two-step expanded format, which might suggest that a non-binary format that includes even more options for gender identity would be even more threatening to those individuals (and rated more negatively).

Study 2: Cisgender Sample #2

This second study served as replication and expansion of Study 1. Based on the results of Study 1, we sought to replicate the confirmation of our previous hypotheses that greater gender binarism, more conservative political orientation, greater religiosity, cisgender heterosexual men, and those with less transgender contact would prefer the binary format and have more negative evaluations of the non-binary formats. We also sought to replicate the finding that distinctiveness threat is associated with preferences for the binary format and more negative attitudes toward the expanded format as well as the new non-binary format included in the present study.

Method

Participants

A total of 371 mTurk workers initially responded. Forty (11%) participants were excluded for failing attention checks. As in Study 1, gender identity of the participants was determined by responses to the expanded format question. A sexual orientation question was added to the demographics section at the end of the survey and used to determine sexual orientation of the participants. An additional 14 participants identified as non-cisgender and 23 identified as LGB, so were excluded from these analyses (non-cisgender and LGB participants from Studies 1 and 2 were merged into a new dataset that was analyzed for Study 3). Additionally, one participant was excluded for responding as cisgender on both the expanded format and demographic items, but one time as male and one time as female.

The final sample size for our analyses was 200 participants (54%). Of those 200 participants, 137 (68.5%) were female and 63 (31.5%) were male, with a mean age of 37.24 (SD = 12.92, range = 18–73). The majority (158, 79%) identified as White (11, 5.5% each Hispanic/Latino, Asian American, and Black; 2, 1% biracial; 1, .5% Native American; 6, 3% other). About 71 (36%) had a high school education or some college, and 129 (64.5%) had an Associate’s degree or higher. The sample was somewhat liberal (92, 46% liberal-very liberal; 62, 31% moderate; 45, 22.5% conservative-very conservative), and 88 (43.8%) were non-religious (112, 56.2% slightly-highly religious).

Procedure and Measures

The design and procedure for this second study was identical to Study 1. All procedures took place online, including consent and debriefing. mTurk workers were paid $.25 for participating. The same measures were used as in Study 1, with a few exceptions. First, an additional non-binary gender questionnaire format was included. Second, four new items were added to address perceptions of institutions that hypothetically used one of the two non-binary questionnaire formats.

Gender Questionnaire Formats

In addition to the two gender questionnaire formats used in Study 1 (i.e., binary format, expanded format), we included the one-item Facebook gender option that lists 58 gender identity choices, but allows for only one choice to be selected. The Facebook format asked “What is your gender? (check one)” and provided the following 58 choices (in order): Agender, Androgyne, Androgynous, Bigender, Cis, Cisgender, Cis Female, Cis Male, Cis Man, Cis Woman, Cisgender Female, Cisgender Male, Cisgender Man, Cisgender Woman, Female to Male, FTM, Gender Fluid, Gender Nonconforming, Gender Questioning, Gender Variant, Genderqueer, Intersex, Male to Female, MTF, Neither, Neutrois, Non-binary, Other, Pangender, Trans, Trans*, Trans Female, Trans* Female, Trans Male, Trans* Male, Trans Man, Trans* Man, Trans Person, Trans* Person, Trans Woman, Trans* Woman, Transfeminine, Transgender, Transgender Female, Transgender Male, Transgender Man, Transgender Person, Transgender woman, Transmasculine, Transsexual, Transsexual Female, Transsexual Male, Transsexual Man, Transsexual Person, Transsexual Woman, Two-Spirit.

Questionnaire Format Attitude Ratings

Participants rated all three formats on the same attitude ratings as used in Study 1. We added three additional items, but these were excluded from analysis due to their redundancy with the original items. We conducted a factor analysis on the attitudes ratings and found that three of the inclusiveness items formed one factor (i.e., inclusiveness) and three formed a second factor (i.e., violation). However, for parsimony, the violation factor was excluded from the analyses (see supplementary materials for all 15 items). The reliability for the inclusiveness items factor was marginally acceptable (binary format: α = .66; expanded format: α = .73; Facebook format: α = .69).

Applied Attitude Ratings

In order to assess the attitudes people hold about hypothetical institutions that utilize non-binary gender questionnaires, we created four items that participants rated on a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). After rating all three of the questionnaire formats individually, participants were asked “How would you feel about a school that used Form B or C on their application?” and “How would you feel about a doctor’s office or medical facility that used Form B or C?” Following each question, participants rated their agreement with the statements: “I would feel that the school [doctor/medical facility] was more welcoming to LGBT students” and “I would be more likely to go to that school [doctor/medical facility] than to one that used a binary gender (e.g., male/female only) question.” Each item was analyzed separately.

Gender Binarism and Distinctiveness Threat

The gender binarism and distinctiveness threat scales were identical to those used in Study 1. Reliability for gender binarism was good in this sample (α = .82). As in Study 1, we used only the first distinctiveness threat item, “Asking about more than two genders (i.e., male and female) threatens the differences between people who are born female and people who are born male” in all analyses, because this item better represents the construct of distinctiveness threat.

Format Preference

Following the format ratings and individual difference variables, participants were asked which of the three formats they preferred. Participants were shown a graphic containing the binary format (labelled “Form A”), the expanded format (labelled “Form B”), and Facebook format (labelled “Form C”). They then were asked, “Which way of asking about gender would you prefer to see on a standard form (e.g., in a research study, for school applications, at a doctor’s office)?,” for which they could select only one format.

Results

Frequency of Format Preferences

Overall, the majority of participants (117, 59.5%) preferred the expanded format over a forced-choice question (69, 34.8% preferred the binary format; 12, 6.1% preferred the Facebook format). Only two participants did not respond to the preference item. A Friedman’s test for ranked totals indicated that there was a significant difference in the ranked preferences for formats (χ 2 = 129.72, p < .001), and median scores showed the expanded format preferred most, the binary format preferred second, and the Facebook format preferred least. Post hoc analyses using Wilcoxon signed-rank tests with a Bonferroni correction (i.e., p = .017) revealed that each signed-rank pair was significantly different from each other: 119 participants ranked the expanded format as their preference compared to the binary format (z = −6.24, p < .001); 118 participants ranked the binary format as their preference compared to the Facebook format (z = −5.85, p < .001); and 146 participants ranked the expanded format as their preference compared to the Facebook format (z = −9.85, p < .001).

Correlations: All Three Formats

Corrected correlations between the questionnaire format attitude ratings and participants’ characteristics (i.e., binary gender, political orientation, age, religiosity, transgender contact) were conducted. See Table 2 for exact values. Perceiving the binary format as more inclusive was associated with higher belief in gender binarism, higher belief that asking about more than two genders threatens the differences between biological sexes, cisgender heterosexual male participants, and more conservative political. Preference for the binary format was associated with greater gender binarism, greater distinctiveness threat, more conservative political orientation, greater religiosity, older participants, and those with less transgender contact. Perceiving the expanded format as more inclusive was associated with lower gender binarism, cisgender heterosexual women participants, and more liberal political orientation. Perceiving the Facebook format as more inclusive was associated with lower gender binarism, cisgender heterosexual women participants, and younger age.

Table 2 Corrected correlations between questionnaire format ratings and participants’ characteristics, and between participants’ characteristics and applied questionnaire format questions, Study 2
Full size table

Applied Rating Items

We also conducted corrected correlations between the applied ratings items (i.e., use of non-binary forms in doctor’s office/medical facility or university) and participants’ characteristics (i.e., binary gender, political orientation, age, religiosity, transgender contact) and the individual difference variables (i.e., gender binarism, distinctiveness threat). See Table 2 for exact values. Cisgender heterosexual women participants perceived doctors’ offices/medical facilities using a non-binary form as more welcoming to LGBT patients. Liberal political orientation was associated with perceptions that universities and medical facilities using non-binary forms would be more welcoming to LGBT persons, and liberal participants would be more likely to attend both universities and medical facilities using non-binary forms. Similarly, younger participants and those with more transgender contact felt using non-binary forms in both scenarios was more welcoming to LGBT people; however, only those with more transgender contact were more likely to attend schools (but not medical facilities) that used non-binary formats. Higher religiosity was associated with less likelihood of attending a university that used a non-binary form on their application. Greater gender binarism and distinctiveness threat were associated with perceptions that the use of non-binary formats in both scenarios was less welcoming to LGBT people, and those higher in gender binarism reported less likelihood of attending universities or medical facilities using non-binary forms.

Logistic Regression

We also conducted a multinomial logistic regression to test the effects of gender binarism, distinctiveness threat, political orientation, binary gender (i.e., female = 0, n = 133; male =1, n = 63), age, transgender contact (no contact = 0, n = 83; contact = 1, n = 113), and religiosity on the likelihood of preferring the binary format over the expanded or Facebook formats (i.e., binary format preference was the reference category). Collinearity statistics indicated that the variables were not problematically correlated with other predictors (i.e., VIF < 10, tolerance > .10; Abu-Bader 2010).

The overall model showed good fit, Pearson χ 2(376) = 357.67, p = .74, and the final model was significantly improved over the intercept-only model (final model likelihood ratio χ 2(14) = 65.14, p < .001; Nagelkerke R 2 = .35). Assessing model fit using the likelihood ratio tests, the nominal variables (i.e., transgender contact, participant binary gender) were not significant. The only significant predictor of binary format preference over the non-binary formats was gender binarism. Those high in gender binarism were roughly twice as likely to prefer the binary format over the expanded format (β = −.82, SE = .28), Wald(1) = 8.35, p = .004, reciprocal odds ratio = 2.27, 95% CI [1.30, 3.97], and approximately five times more likely to prefer the binary over the Facebook format (β = −1.62, SE = .59), Wald(1) = 7.45, p = .006, reciprocal odds ratio = 5.05, 95% CI [1.58, 16.13].

Discussion

As in Study 1, attitudes about gender questionnaire formats were most strongly associated with gender binarism beliefs, distinctiveness threat, participants’ gender, and political and religious ideologies. Although it is important to understand the potential motivation behind gender questionnaire format preferences among cisgender people (who tend to be the ones making decisions about questionnaire formats), of greater importance is understanding the preferences of non-cisgender individuals. Studies 1 and 2 examined how cisgender individuals felt about the traditional binary gender questionnaire format compared to non-binary gender formats, but we were interested in investigating the opinions of non-cisgender participants about how researchers, businesses, and the medical profession ask about and classify gender.

Study 3: LGB and Gender-Diverse Samples

The format for gender questions is hotly contested in the real world, with proponents of expanded measures arguing for the importance of allowing gender-diverse persons to self-identify for accuracy of data (Cahill et al. 2014; Deutsch et al. 2013) and for feelings of inclusion (GenIUSS 2014; Tate et al. 2013). However, those who are most affected by the gender question format (i.e., transgender and non-binary individuals) are often not given a chance to express their opinions on the matter of how gender identity is asked on questionnaires. Rather, it is more common that cisgender persons make assumptions about what is best for the transgender and gender non-binary communities. In this third study, we sought to sample transgender and gender non-binary individuals to ask them about their opinions. We were also interested in how cisgender LGB individual’s opinions about gender questionnaire formats compared to the attitudes of the cisgender samples of Studies 1 and 2 and compared to the attitudes of transgender and gender non-binary individuals. We had one research question: What are the opinions of transgender and gender non-binary, and other LGB individuals regarding the presentation of gender categories? We were still interested in how individual difference variables (i.e., gender binarism, distinctiveness threat, political orientation, religiosity, age, gender, and transgender contact) affected format preferences, but these relationships of interest remained exploratory because research on the attitudes of sexual minority individuals toward gender-diverse individuals is limited.

Method

Participants

We recruited a LGBT sample from LGBT centers in the Midwestern United States, Facebook, and mTurk. As part of the recruitment from LGBT centers and Facebook, participants could enter a lottery for one of four $25 Amazon gift cards that would be randomly awarded at the end of data collection. Of the 29 LGBT centers contacted, eight agreed to forward our invitation to their members. The survey sent to the LGBT centers and Facebook groups was the same as the survey used in Study 1. Because the participation rates from the LGBT centers and Facebook groups were extremely low, we also recruited LGBT participants through mTurk. We posted a research participation opportunity (and survey link) that was limited to individuals who “identify as non-cisgender or non-heterosexual” and included screening questions for eligibility. Participants were compensated $.25 for their participation. The survey used for this sample was the same as the survey used in Study 2.

Finally, we merged the data from the LGBT centers and Facebook groups (n LGB = 9; n non-cisgender = 11), the LGBT mTurk sample (n LGB = 104; n non-cisgender = 55), and the LGB and non-cisgender participants who participated in Studies 1 and 2 (but were excluded from analyses in those cisgender samples; n LGB = 24, n non-cisgender = 20). Because we were interested in comparing the attitudes of LGB and non-cisgender individuals, we split the merged data into two separate analyses based on gender identity (i.e., cisgender LGB vs. non-cisgender across sexual orientations, as determined by responses to the expanded format items).

A total of 137 LGB participants were included in this sample (M age  = 30.90, SD = 10.62, range = 19–66) and all responded to the demographic questions. All identified as cisgender and 101 (73.7%) identified as female (male = 36, 26.3%). All participants identified their sexual orientation as bisexual (82, 59.9%) or gay/lesbian (55, 40.1%). The majority of the LGB sample was White (121, 88.3%); 6 (4.4%) Black, 3 (2.2%) each Hispanic/Latino or Asian, 3 (2.2%) biracial, and 1 (.7%) Native American. The sample was educated (71, 51.8% with a Bachelor’s degree or higher), very liberal (108, 78.9% liberal-very liberal; 9, 6.6% conservative-very conservative), and less religious (92, 67.2% not religious).

The total number of transgender and gender non-binary participants was 86 (M age  = 29.93, SD = 9.99, range = 18–64) and between 88 and 98% (depending on the question) responded to the demographics questions. The majority (86, 64%) identified their gender as one of the gender-diverse options (i.e., Female-to-Male, Male-to-Female, genderqueer) or “additional gender category/other” (24, 27.9%; text entries included agender, androgynous, avatrix, gender fluid, non-binary, pangender, and gendervague), with 4 (4.7%) identifying as male and 3 (3.5%) identifying as female. Most of the transgender and gender non-binary participants identified their sexual orientation (11, 12.8%, participants did not respond to this question) as gay/lesbian (12, 16%), bisexual (22, 29.3%) or other (36, 48%; text entries included ACE spectrum, asexual, demisexual, pansexual, queer, polysexual, and fluidsexual); only five participants (6.7%) identified as heterosexual. The majority of the sample was White (69, 82.1%); 5 (6.0%) Hispanic/Latino, 2 (2.4%) each Black, Asian, and Native American, 3 (3.6%) biracial, and 1 (1.2%) other. Forty-three percent (36 participants) had a Bachelor’s degree or higher, the vast majority were liberal (71, 84.6% liberal-very liberal; 3, 3.6% conservative-very conservative), and most were not religious (49, 58.3%).

Procedure and Measures

LGBT participants were recruited from U.S. Midwestern LGBT centers, large U.S.-based Facebook groups, and through mTurk. Administrators of the LGBT centers and Facebook groups were contacted through private emails asking if they would be willing to forward or post online the invitation to participate (including a study description, contact information, and a link to the online survey) to members of their group. LGBT participants were recruited via mTurk by posting a research participation opportunity (and survey link) with the eligibility requirement of being LGBT. All participants accessed the online survey via a hyperlink. The online survey procedures were identical to those used in Studies 1 and 2.

A majority of the sample responded to a survey identical to that used in Study 2. Six participants who identified as transgender participated in Study 1 and 20 participants who were recruited from LGBT centers and Facebook groups completed the same survey as in Study 1 (with an added demographic question for sexual orientation).

Questionnaire Format Attitude Ratings

The same three gender questionnaire formats used in Study 2 were used in Study 3, with the exception of the 26 participants who completed the survey from Study 1 (i.e., only saw the binary and expanded formats, not the Facebook format). The same inclusiveness factor used in Study 2 was used in the analyses for this third study. (As in Study 2, we excluded the violation factor from analyses due to redundancy with other variables.) Reliabilities in the LGB sample ranged from acceptable to good for the inclusiveness items: the binary format, α = .81; expanded format, α = .82; the Facebook format, α = .80. Reliabilities in the transgender and gender non-binary sample ranged from acceptable to good for the inclusiveness items: the binary format, α = .86; expanded format, α = .85; the Facebook format, α = .87.

Applied Attitude Ratings and Format Preference

The same items for attitudes toward universities and doctor’s offices/medical facilities that used non-binary gender questionnaires as in Study 2 were used. Each item was analyzed separately. Format preference was assessed using the same questions and graphic as in Study 2. Participants selected only one format as their most preferred.

Gender Binarism and Distinctiveness Threat

The gender binarism and distinctiveness threat scales were identical to those used in Studies 1 and 2. Reliability for the gender binarism scale was good in the LGB sample (gender binarism: α = .81) and in the transgender and gender non-binary sample (gender binarism: α = .81), but the reliabilities were unacceptable for distinctiveness threat in both samples (LGB: α = .46; transgender and gender non-binary: α = .39), so again, only the first distinctiveness threat item was used in all analyses.

Results: LGB Sample

Frequency of Format Preferences

The nine LGB participants from the LGBT centers and Facebook-recruited sample were not exposed to the Facebook format, and so they were excluded from analyses of format preference (including the multinomial regression), leaving an effective sample size of 128. The majority of participants preferred the expanded format (97, 75.8%), followed by the binary format (18, 14.1%), and then the Facebook format (13, 10.2%). A Friedman’s test for ranked totals indicated that there was a significant difference in the ranked preferences for formats (χ 2 = 106.73, p < .001), and median scores showed the expanded format was preferred most, the Facebook preferred second, and the binary format preferred least. Post hoc analyses using Wilcoxon signed-rank tests with a Bonferroni correction (i.e., α = .017) revealed that the expanded format was preferred over the binary format (z = −8.28, p < .001) and over the Facebook format (z = −7.64, p < .001). There was not a significant difference in ranks between the Facebook and binary formats (z = −2.09, p = .04, with the Bonferroni corrected p = .017).

Correlations: Format Evaluations

Corrected correlations were conducted between the questionnaire format attitude ratings and participants’ characteristics (i.e., binary gender, political orientation, age, religiosity, transgender contact. See Table 3 for exact values. As in the cisgender samples from Study 1 and 2, conservative political orientation was associated with perceptions of the binary format as more inclusive. Age was associated with attitudes toward the gender formats such that older LBG participants had a stronger preference for the binary format and perceived the binary format as more inclusive.

Table 3 Corrected correlations between gender questionnaire format attitudes ratings, participants’ characteristics, and individual difference measures, Study 3
Full size table

We also conducted corrected correlations between the questionnaire format attitude ratings and the individual difference variables of gender binarism and distinctiveness threat (see Table 3). Compared to the cisgender samples (Studies 1 and 2), mean scores were slightly lower for gender binarism and slightly below the scale midpoint (i.e., 2.5), and the distinctiveness threat item mean was slightly lower compared to the other cisgender samples. As in the cisgender samples, greater gender binarism beliefs in the LGB sample were associated with perceptions that the binary format was more inclusive and with strong preference for the binary format. Conversely, low gender binarism was associated with perceptions of the expanded and Facebook formats as more inclusive. Distinctiveness threat also showed similar patterns in the LGB sample as in the cisgender samples: higher belief that asking about more than two genders threatens differences between biological sexes was associated with perceptions of the binary format as more inclusive, perceptions of the expanded and Facebook formats as less inclusive, and a strong binary format preference.

Applied Ratings

Corrected correlations were also conducted between participants’ characteristics, individual difference variables, and the applied questions which assessed perceptions of universities and medical facilities that utilize non-binary gender question formats. See Table 4 for exact values. Perceptions that institutions using non-binary questionnaire formats were less welcoming to LGBT members, and less likelihood of patronizing universities or medical facilities using non-binary formats was associated with greater gender binarism, cisgender male participants, and more conservative political orientation. Greater belief that asking about more than two genders threatens the biological differences between sexes was associated with perceptions of both institutions as being less welcoming and with less likelihood of patronizing those institutions. Older participants were associated with perceptions that a medical facility using a non-binary format would be less welcoming to LGBT clients.

Table 4 Corrected correlations between participants’ characteristics and applied questionnaire format questions, Study 3
Full size table

Logistic Regression

A multinomial logistic regression was conducted for the LBG sample only. As in the previous analyses, we excluded nine participants from the LGBT centers and Facebook-recruited sample who only viewed two of the formats, not all three.

We conducted a multinomial logistic regression to test the effects of political orientation, binary gender (i.e., cisgender female = 0, n = 96; cisgender male = 1, n = 32), age, transgender contact (no contact = 0, n = 20; contact = 1, n = 108), and religiosity on the likelihood of preferring the binary format over the expanded or Facebook formats (i.e., binary format preference was the reference category). Collinearity statistics indicated that the variables were not problematically correlated with other predictors (i.e., VIF < 10, tolerance > .10; Abu-Bader 2010). Gender binarism and distinctiveness threat were not included in the model because these variables are likely too hetero- and cis-normative to be a meaningful construct in this sample (as evidenced by the low reliability for these items and the low mean scores on these constructs).

The overall model showed adequate goodness-of-fit, Pearson χ 2(226) = 246.92, p = .16, and the final model was significantly improved over the intercept-only model (final model likelihood ratio, χ 2(10) = 25.63, p = .004 (Nagelkerke R 2 = .24), suggesting that at least one variable in the regression is greater than zero. Of the nominal variables (i.e., transgender contact, participant binary gender) only having transgender contact had an overall significant effect, χ 2(2) = 5.94, p = .51. Those with transgender contact were nearly five and a half times more likely to prefer the expanded format over the binary format (β = −1.69, SE = .68), Wald(1) = 6.21, p = .01, reciprocal odds ratio = 5.43, 95% CI [1.43, 20.83]). However, transgender contact did not predict preference for the Facebook format over the binary format (β = −1.65, SE = .95), Wald(1) = 4.07, p = .16. Conversely, older participants were more likely to prefer the binary format over either of the lengthier formats—expanded: β = −.08, SE = .03, Wald(1) = 10.19, p = .001, reciprocal odds ratio = 1.09, 95% CI [1.03, 1.14]; Facebook: β = −.16, SE = .06, Wald(1) = 7.55, p = .006, reciprocal odds ratio = 1.17, 95% CI [1.05, 1.31].

Results: Gender-Diverse Sample

As in the LGB sample, participants from Study 1 and from the LGBT centers (who did not see the Facebook format) were excluded from analyses, leaving an effective sample of 69 respondents. Most participants preferred the expanded format (42, 62.7%) or the Facebook format (22, 32.8%); only three participants (4.5%) preferred the binary format. A Friedman’s test for ranked totals indicated that there was a significant difference in the ranked preferences for formats (χ 2 = 54.97, p < .001), and median scores showed the expanded format preferred most, the Facebook format preferred second, and the binary format preferred least. Post hoc analyses using Wilcoxon signed-rank tests with a Bonferroni correction (i.e., α = .017) revealed that each signed-rank pair was significantly different from each other: the expanded format was preferred over the binary format (z = −6.30, p < .001) and over the Facebook format (z = −3.03, p = .002); the Facebook format was preferred over the binary format (z = −4.72, p < .001).

Correlations: Format Evaluations

We conducted corrected correlations between relevant questionnaire format attitude ratings and participants’ characteristics (see Table 3). Among transgender and gender non-binary participants, political orientation was associated with attitudes toward the gender questionnaire formats. More conservative participants had similar perceptions of the questionnaire formats as conservative cisgender participants: greater perceived inclusiveness of the binary format and less inclusiveness of the expanded format. Not surprisingly, the average level of gender binarism and both distinctiveness threat items were lower in the transgender and gender non-binary sample than in the cisgender samples. However, greater gender binarism was associated with perceptions of the binary format as being more inclusive and preferences for the binary format. Greater distinctiveness threat was associated with binary format preferences.

Applied Ratings

We conducted corrected correlations between the applied questions (i.e., attitudes about universities and medical facilities using non-binary formats) and participants’ characteristics and individual difference measures (see Table 4). Only more liberal political orientation was associated with perceptions that institutions using non-binary questionnaire formats would be more welcoming to LGBT members and more likelihood of patronizing schools using the non-binary forms, but not medical facilities). A multinomial logistic regression for the transgender/non-binary sample could not be run because the number of observations was less than the number of specified parameters.

Discussion

As in the other cisgender samples (Studies 1 and 2), attitudes held by cisgender (LGB) participants about gender questionnaire formats were most strongly influenced by participants’ transgender contact and age. Although the majority of the LGB sample reported a preference for the expanded format, older age predicted preferences for the binary format. Almost all of the transgender and gender non-binary participants reported a preference for the expanded format as well, followed by a preference for the Facebook format, with only three non-cisgender participants selecting the binary format as their preference.

General Discussion

Across three studies, including one with exclusively LGBT participants, we found consistent support for our predictions that multiple individual difference variables (i.e., gender binarism, distinctiveness threat, political conservatism, older age, cisgender male participants, religiosity, low transgender contact) associated with anti-transgender attitudes were also associated with binary format preferences. Our exploratory relationship of interest, distinctiveness threat, was shown across all three studies with cisgender participants to be associated with preference for the binary format, suggesting that those with greater distinctiveness threat may be threatened by the mere suggestion that more than two genders exist. Additionally, these individual difference variables were associated with perceptions of institutions of higher education or medicine that used non-binary question formats on official forms as less welcoming to LGBT individuals and lower likelihood of patronizing such institutions.

Conversely, participants low in gender binarism and distinctiveness threat, more liberal, less religious, younger, and with more transgender contact consistently rated the expanded format as more inclusive and showed a strong preference for the non-binary formats. Among the gender-diverse sample of participants, we found a significantly greater preference for the expanded format over either the Facebook or binary formats. Together, these findings may suggest that opposition to the implementation of non-binary gender question formats may be based on implicit negative attitudes about gender diversity and non-cisgender persons rather than on concerns about the privacy or safety of gender-diverse individuals.

In Study 1, using a cisgender sample, only two formats were presented: the binary format and expanded format. Overall, gender questionnaire format preferences were about evenly split in this sample, although preference for the binary format was correlated with more conservative political orientation, greater gender binarism beliefs, greater distinctiveness threat, higher religiosity, and less transgender contact. In Study 2, where a second non-binary format was included, the majority of participants rated the expanded gender questionnaire format as the preferred format as compared to the traditional binary and Facebook formats, although the binary format was ranked significantly better than the Facebook format. Non-binary gender questionnaire format preference was correlated with liberal political orientation, younger participants, and those with more transgender contact. As in Study 1, preference for the binary format was associated with conservative political orientation as well as with older, more religious participants and with those high in gender binarism and distinctiveness threat beliefs. Indeed, the likelihood of preferring the binary format over the expanded format was predicted by high gender binarism in both cisgender samples.

Although no known research to date has directly assessed format preferences, many of the variables associated with binary format preference among cisgender participants are consistent with variables implicated in prior research as associated with negative attitudes toward transgender persons and anti-trans prejudice. Anti-trans attitudes have been linked to greater gender binarism, political conservatism, and religiosity or religious fundamentalism and were significantly higher among cisgender heterosexual men than cisgender heterosexual women (Carroll et al. 2012; Hill and Willoughby 2005; Nagoshi et al. 2008; Norton and Herek 2013; Tebbe and Moradi 2012). These same variables tended to be associated with binary format preference in our cisgender samples, which may suggest that less favorable attitudes toward the expanded formats is related to negative attitudes toward transgender individuals. It is possible that those who have more negative attitudes toward transgender individuals are more opposed to non-binary gender questionnaires because they do not want to acknowledge or normalize non-binary gender identities.

Study 2 also included applied questions, asking participants to rate their attitudes toward a university or a medical facility that used a non-binary gender form. Cisgender heterosexual women and more liberal participants rated the university and medical facility using non-binary gender forms as more welcoming to the LGBT community and reported they would be more likely to patronize such establishments. Conversely, those high in gender binarism and distinctiveness threat rated both universities and medical facilities using non-binary formats as less welcoming and reported they would be less likely to patronize such establishments. These findings may best illustrate the proposed connection between transgender attitudes and attitudes toward gender questionnaire formats: participants high in gender binarism and who are more politically conservative (variables associated with anti-trans prejudice; Hill and Willoughby 2005; Norton and Herek 2013; Tebbe and Moradi 2012) reported that they would not patronize institutions using expanded format questionnaires, whereas cisgender heterosexual women and liberals (variables associated with less anti-trans prejudice; Norton and Herek 2013; Tebbe and Moradi 2012) reported more favorable attitudes toward institutions implementing expanded questionnaire formats.

Study 3 had two main purposes: to assess LGB participants’ attitudes regarding gender questionnaire formats in relation to other cisgender samples and the transgender and gender non-binary participants, and, more importantly, determine the attitudes and preferences of gender non-binary participants (as they are the people most affected by the formats used). LGB (cisgender) participants’ attitude ratings were quite similar to those of the other cisgender samples. Gender binarism and distinctiveness threat were associated with perceptions of more inclusiveness of the binary format and preferences for the binary format.

Although gender binarism and distinctiveness threat showed the same correlational patterns in the transgender and gender non-binary sample, mean levels of both measures were very low. More importantly, for those individuals for whom gender questionnaire formats have the greatest personal impact, the binary format was rated as the least preferred, with the expanded format ranked as most preferred. The trends that cisgender-identified heterosexual participants show the least support for non-binary formats, as well as that transgender and gender non-binary participants show the most support for non-binary formats—with non-heterosexual cisgender participants somewhere in the middle—may reflect the benefits of the inclusive items for those participant’s groups. Whereas the heterosexual cisgender participants may only benefit from inclusion indirectly, transgender and non-binary people will benefit directly. Because non-heterosexual and non-cisgender persons are often grouped together (such as LGBT), it may be that the benefits of using non-binary formats are more obvious and relatable to the non-heterosexual cisgender sample as well.

Interestingly, across all three studies, participants—even those who endorsed the non-binary formats in general—tended to dislike the Facebook format. It is possible that 58 options for gender identity and expression is overwhelming, especially for cisgender participants (who are used to having their gender identity “box” front and center as the first option). Additionally, the Facebook format only allows for the selection of a single gender identity (presumably as a trade-off for the substantial number of identities), which may be too restrictive for certain people who would prefer to select multiple identities that they feel suit them. Furthermore, the Facebook format shares another potentially negative aspect with the binary format: there is no option for “other” or for a write-in response. Many of the transgender and gender non-binary participants in our study self-identified using the “other” option and writing in their own preferred gender identity nomenclature. It is possible that, especially among non-cisgender individuals, the lack of open-text response options and single-choice response makes the Facebook format less desirable. It should be noted that in 2014, Facebook altered their gender question format to be more inclusive by allowing individuals to select up to ten gender identity labels from the list and also added a “custom” option which allows people to write in their own identity label (Bell 2015; Facebook Diversity 2015; Ferraro 2014; Griggs 2014; Williams 2014).

Our results support the arguments presented in the literature in favor of implementing non-binary gender questionnaire formats (e.g., Cahill et al. 2014; Deutsch et al. 2013; GenIUSS 2014; IOM 2013; Tate et al. 2013). Cisgender, LGB, and transgender and gender non-binary participants who prefer the non-binary formats consistently reported that they felt the non-binary formats were more inclusive, and in applied settings (i.e., university admissions, medical facilities), institutions using non-binary formats were seen as more welcoming to the LGBT community and more likely to be patronized by participants low in gender binarism beliefs and distinctiveness threat, as well as by transgender and gender non-binary participants.

Limitations and Future Directions

Our studies are not without limitations. Foremost, our studies were correlational and did not assess reactions to actually using different formats (e.g., felt and internalized stigma, Herek 2007; minority stress, Frost and Meyer 2009; Meyer 2003). Although our studies were not experimental, nor did they include specific theoretical hypotheses regarding stigmatization, based on the overwhelming preference for the non-binary questionnaire formats, it seems that the transgender and gender non-binary participants in our sample would like the option to express their gender identity. Still, the questions of whether they choose to do so or not and for what reasons would benefit from further experimental study. Additionally, we only used two formats for non-binary questionnaires. It is possible that other non-binary formats could be preferred by transgender and gender non-binary individuals. Future research may assess attitudes toward gender questionnaire formats that include birth sex and gender expression of appearance and mannerisms (e.g., Wylie et al. 2010), or formats similar to our expanded format, but with additional, specific non-binary options for gender identity (e.g., Reisner et al. 2014).

Although it is beyond the scope of the current studies, which focused on people’s perceptions of and attitudes toward gender question formats, future research should further address the psychometric properties of different gender question formats (i.e., beyond missing data and response rates). An understanding of which formats provide the most accurate and reliable data is also important when considering best practices for collecting data on sex and gender identity.

Practice Implications

Based on the findings of our studies, researchers, practitioners, and institutions seeking to be inclusive or to identify gender-diverse individuals should use the expanded (two-question) format, especially over the binary or Facebook formats. The present studies assessed attitudes toward several different formats and participants’ preferences show that among LGBT participants, the expanded format was preferred and the binary format was not. Among cisgender heterosexual participants, the binary format was also never ranked significantly higher than the expanded format; indeed, in Study 2, the expanded format was ranked significantly higher than the binary format. These findings likely indicate a few practical implications: (a) that gender-diverse individuals would prefer to be able to identify themselves if they want, which requires the use of a non-binary format gender question; (b) even among cisgender individuals, many prefer the expanded format because they feel it is more inclusive to everyone; and (c) cisgender participants who may be offended by non-binary formats appear to share some specific social identities (e.g., conservative, religious, older, cisgender men) and beliefs (e.g., that there are only two genders and sexes, asking about more than two genders threatens differences between sexes).

Additionally, although cisgender respondents tended to prefer the binary format over the expanded format, prior research has shown behaviorally that cisgender respondents will still use the “expanded” (i.e., two-item: gender identity and sex assigned at birth) question with a near 0% missing data rate (Tate et al. 2013). As previously discussed, prior research has also assessed the validity of question options for correctly identifying non-cisgender individuals and found that the two-step approach is the most effective, especially because some transgender people identify as male or female, not as transgender male or transgender female (Tate et al. 2013).

In conjunction with our recommendation to use non-binary format gender questions, several recent publications have endorsed the importance of including more than binary options for sex and gender in academic and medical research and records (e.g., GenIUSS 2014; IOM 2013; Runnels et al. 2014; Westbrook and Saperstein 2015) and even in non-academic settings such as marketing and social media (e.g., Dholakia 2016). A common idea for consideration is the write-in option for gender identity, which would potentially allow gender-diverse and cisgender individuals to self-identify as generally or specifically as they please, without having to fit into a pre-determined gender category. However, in practical terms, completely open-ended questions for gender identity invite methodological and data quality problems such as the time and resources needed to code and analyze qualitative data and the possibility that people will intentionally or unintentionally provide meaningless responses (Monette et al. 2013). One compromise suggested is to provide a write-in option for those who select a non-binary gender identity only; this would allow for gender-diverse individuals to provide more accurate and specific information if they choose, while keeping the amount of qualitative data to analyze relatively small (Dholakia 2016; GenIUSS 2014).

Conclusion

The topic of how gender identity data is collected continues to be debated, particularly in situations where institutions vote on changing from the traditional binary format. Our research examined the attitudes of cisgender heterosexual, LGB, and gender-diverse participants toward binary and non-binary gender question formats to assess what people would prefer to see used by institutions and researchers, and it also identified social identities and beliefs that might underlie oppositional attitudes toward more inclusive measures of gender identity. Across three studies, a majority of participants rated the expanded format for measuring gender identity as the most preferred format. Preferences for the standard, binary format were associated with and predicted by beliefs associated with prejudice toward transgender and gender non-binary individuals. These studies provide strong evidence that (a) most cisgender people prefer the expanded questionnaire format and (b) that gender-diverse individuals would prefer the option to “check their box” using the expanded format.