Qualitative research on gender identity: phase 1 summary report

We have conducted qualitative research to help inform whether and how best to meet user needs for information on gender identity – the main one being a reliable estimate of the population identifying as “trans” (an umbrella term covering people who self-identify as an identity different to the sex assigned to them at birth).

We explored relevant issues, including responses to sex and gender questions and potential barriers to answering, terminology, privacy, burden and acceptability. The Office for National Statistics (ONS) Data Collection Methodology branch undertook 4 focus groups and 18 one-to-one in-depth interviews with members of the public, including those with a trans identity and those who are cisgender (people whose self-identity conforms to the sex or gender assigned at birth). These were conducted in March and April 2017.

Three question designs were explored:

• the 2011 Census “Sex” question
• a hybrid design with the addition of “Other” to the sex question
• a two-step design with separate sex and gender identity questions

None of the question designs, as they were presented, would meet the requirement for a reliable estimate of the trans population. Furthermore, none of the question designs would fully meet respondent needs for questions that are easily understood and answered. It is therefore recommended that none of these designs be used in the 2021 Census.

Our 2021 Census topic consultation identified a need for information on gender identity. As part of our work to help determine whether, and how best, to meet user needs for information on this topic, the ONS Data Collection Methodology branch was commissioned to undertake qualitative research. This work is in addition to other strands of research and testing we have committed to in our Gender identity research and testing plan. This report details Phase 1 of the qualitative research. It covers: the purpose of the research, the methodology used, as well as findings, conclusions and recommendations. We will be undertaking Phase 2 of the qualitative research in due course.

The objectives were to provide findings and recommendations with regard to: how the trans population tackle the 2011 Census “Sex” question, including potential barriers to answering the question and completing the census; how both the trans population and cisgender population (people whose self-identity conforms to the sex or gender assigned at birth) interpret concepts around gender identity and sex; how both populations might answer questions on sex and gender identity; what terminology respondents understand and enables them to answer the question(s) appropriately; and potential privacy, security, burden and acceptability concerns.

A purposive sample design was employed. The primary sampling criterion was sex or gender. Four focus groups were conducted with the cisgender population (total 29 participants) and 18 one-to-one in-depth interviews were conducted with the trans population, including trans men, trans women, non-binary and intersex identities, and the parents of trans children. Secondary criteria included age, educational attainment, ethnic group and household type (see Appendix A: Achieved sample). Note that it is not appropriate or possible to draw statistical inferences from the findings when purposive sampling methods are used.

A topic guide covered the areas to be probed, with some tailoring for the cisgender group, the trans group and the parents of trans children. Participants were shown three different question designs (detailed in this section). For each design they were asked for their initial views; their understanding of the questions, the measurement concepts and related terms; whether and how they might answer them; their reaction to each design potentially being used in the census; and suggestions for changes. Probing was also conducted about aspects of the census, such as questions being mandatory or voluntary, data security and confidentiality, privacy within the household, answering by proxy, and the purpose of such questions.

The three question designs explored with all participants:

1. The 2011 Census question: ‘What is your sex?’ 1. Male, 2. Female

2. A “hybrid” question, still asking ‘What is your sex?’ but which added a third response category, “Other” potentially allowing varied interpretation of sex or gender

3. A two-question design:

Question 1: ‘What is your sex?’ 1. Male, 2. Female

Question 2: ‘Which of the following options best describes how you think of your gender identity?’ 1. Male, 2. Female, 3. In another way

The three question designs Meeting data requirements

None of the question designs, as they were presented, would meet the main data requirement of a reliable estimate of the trans population:

• the 2011 Census question provides no means of distinguishing trans and cisgender people
• the hybrid question enables respondents to answer “Other” but a proportion of trans people would answer “Male” or “Female”, so it would undercount the trans population
• the two step question design would not necessarily result in answers that would identify a trans person; so again, there would be an undercount of the trans population

Question validity

None of the questions are completely valid, that is conceptually clear and consistently understood and answered by respondents. This would result in some measurement error. From qualitative research we cannot know the extent of error; at the whole population level it is probably very marginal, but might be more significant with regard to the population of interest.

Cisgender participants could answer the 2011 “Sex” question and the hybrid question without error. However, interpretations and therefore answers that would be given varied among the trans population.

Trans participants expressed confusion about the subject of the “Sex” question; either thinking the “Sex” question referred to biological sex, or that it was indirectly asking about gender identity. Either way, they expressed negativity about the question and confusion about how to answer.

For the hybrid question, the retention of the word “sex” in the stem paired with “Other” in the options caused trans participants to be unsure whether they were being asked about their biological sex (when “Other” would be intersex) or their gender (when “Other” would be open to them). This led to inconsistencies in responses.

The two step design was something of an improvement but again was met with varied interpretations and answer strategies. The distinction between the concepts of sex and gender was broadly recognised but the need for two questions and the distinction between them was not always clear. Thus there arose some confusion, uncertainty in answering and indications that Question 2 might sometimes be skipped, including among cisgender people.

Meeting respondent needs

Furthermore, none of the question designs would fully meet respondent needs for questions that are easily understood and answered, inclusive of all population groups and acceptable.

The 2011 “Sex” question was considered to be irrelevant, unacceptable and intrusive, particularly to trans participants, due to asking about sex rather than gender. The lack of response options for intersex and non-binary people could make it difficult to answer and cause feeling of having to make a forced choice, leading to potential item non-response or invalid response, non-response to the census as a whole and could cause complaints to ONS.

The hybrid design was thought an improvement. However, asking about sex was again thought irrelevant and intrusive. The addition of “Other” caused respondent uncertainty as to whether the question was actually about gender. The “Other” category was thought to homogenise trans people and differentiate them from the rest of society. A “specify” field would mitigate the negative views to some extent.

The two step design was considered even more of an improvement, because Question 2 introduced gender to the equation and the category “In another way” allowed expression of non-binary gender and other trans identities. However, the same feelings arose for Question 1 as did with the 2011 question. The wording of Question 2 met with some differences of opinion as to the use of the word “identity”. There were cisgender and trans participants who did not understand why there were two questions, leading to uncertainty in answering and potentially measurement error or item non-response to Question 2.

Trans participants had mixed feelings that their two answers, in combination, might or might not result in their trans identity being visible in the census data. Some were positive that they could be represented in statistics about trans people, which they felt was important. However, others expressed negativity, either because if they gave the same answers to the two questions their trans identity would not be visible in data (at the aggregate level) when they wanted it to be; or, conversely, if their answers differed they would be visible in the trans statistics, when they did not want to be visible. They did not feel the questions were clear enough, in meaning and purpose, to enable an informed decision as to how to answer.

Conclusions and recommendations

We recommend that none of the three designs be used in the 2021 Census. Even if it is not possible to meet data requirements, change should be made to better meet the needs of trans respondents (for example, removing “sex” and adding one or more additional categories for non-binary and intersex people, recognising that those are distinct).

It is not possible to firmly recommend specific changes to question(s) because data requirements should first be clarified (including for the 2011 “Sex” question). Which is or are the appropriate measurement concept(s) – for example, sex assigned at birth, transitioned sex, reclaimed sex, acquired gender, gender or gender identity – is a matter for consideration by ONS and data users. Stakeholders should be informed of these findings and the implications for what data can be collected.

Any changes need to consider legislative and ethical issues. For example, to meet the data requirement questions on sex at birth and current sex or gender would be needed but might be precluded by the Gender Recognition Act and the discomfort felt among trans people for such questions, with implications for the mandatory status of one or other question.

Changes to questions might have further implications both in the census context (editing and imputation procedures; linking with administrative data) and more widely (the National Statistics Harmonised Standard and social survey data collection – for example, the processing of derived variables and suitability of questions for interviewer modes).

Consideration could be given to approaches used in other countries, such as Australia, where only male and female response options are provided but another means of expressing a trans identity is available.

Consideration should be given to provision of explanation as to why a particular design has been chosen, the meaning of question(s) and guidance on answering, within the census. However, we caution against over-reliance on guidance, given the typically low rates of referral by respondents.

Various specific recommendations are made relating to the different designs presented to participants and to potential redesigns, relating to the question stems, existing or additional response categories and the presentation of questions on paper and online. They are too many and detailed, and too contingent on other factors outlined previously, to be summarised concisely here.

Any changes made should be thoroughly tested using qualitative and quantitative methods, to ensure designs would better meet data requirements and respondent needs without unforeseen negative consequences, such as impact on either trans or cisgender people’s comprehension, ability to answer and the acceptability of questions.

General acceptability of the topics of sex and gender

There is sufficient positive evidence from the Phase 1 research to recommend that work continues to consider inclusion of questioning about gender or trans identity for the 2021 Census or social surveys. Broadly speaking, inclusion of the topic of gender was acceptable across cisgender and trans groups. However, there were objections too. The detail of the questions asked was central to views about the acceptability of inclusion in the census or social surveys. As well as the question designs themselves, other aspects of the census influenced participants’ views of the acceptability of inclusion and the likely accuracy of the responses that would be provided.

Scope – of whom should questions be asked?

No subgroups should be excluded from being asked about sex and/or gender due to considerations of either relevance or acceptability. However, consideration should be given to what the appropriate age to start asking about the gender (identity) of children is.

Mandatory versus voluntary questions

Decisions as to which question or questions might be mandatory and which voluntary must take into account several factors: acceptability to respondents, the potential impact forced response might have on data quality and Census unit response and legal considerations.

We would tentatively recommend that an unchanged 2011 Census question should not be mandatory, for the benefit of, particularly, intersex and non-binary people who cannot choose male or female as a reflection of their current sex or gender. We would also recommend that a question specifically about sex at birth, or one that forces an answer indicating a trans identity, should not be mandatory given the needs of trans people living with disclosure concerns. Perhaps as long as one or other question has a substantive answer, that would be acceptable. However, “Other” or “In another way” type categories could potentially be answered without a valid sex or gender, which would result in unusable data.

Purpose and trust

The views of participants about the purpose of the question(s), trust in the data users or uses and in data security and confidentiality, lead us to recommend consideration of: whether general census guidance is adequate to meet the need for information and reassurance participants said they required, in conjunction with the provision of question-specific guidance; requirements relating to census publicity and to the Census helpline in preparing the ground; and concerns about personal data going into the public domain after 100 years.

Government social surveys

The following points should be considered in assessing whether to take forward collection through government social surveys rather than (or as well as) the census. The issues regarding the validity and unacceptability of the 2011 “Sex” question suggest its continued use as a harmonised standard should be reviewed. Other implications of change would need to be managed: for example, relevance of questions to the specific survey topic; the complexity of survey systems that refer to the “Sex” variable; the current practices of interviewers in asking or recording a respondent’s sex; and methods that maintain a respondent’s privacy. Further development and testing would be required to introduce revised questioning on social surveys.

The varied approaches to completing the individual sections of the census questionnaire taken within households – individually, collectively, personally and by proxy – have implications for the accuracy of data about sex and gender.

Within-household privacy

Cisgender participants had no concerns about their responses to questions being known to other household members, when each individual’s questions were answered in a single online or paper questionnaire. Trans participants varied as to whether they shared this view. Some indicated that they would be uncomfortable and may give answers they deemed to be socially desirable instead. This would result in an undercount of the trans population; of what extent we cannot say.

Proxy data collection

Views about proxy response were varied. There were participants in both the cisgender and trans groups who would be happy to answer sex and/or gender questions on behalf of others in their households, or to have their questions answered by others. However, there were also views against, on the basis of it being a person’s right to answer personal questions for themselves and because the answer might not be accurate.

Although there is potential for complaints to be made if respondents are unhappy that their questions have been answered by proxy without their consent, pragmatically the census can’t disallow proxy response without increasing item non-response to levels that would severely compromise data utility. The likelihood of thus undercounting the trans population should be explained to stakeholders.

”Prefer not to say” or similar response options

It is recommended that “prefer not to say” (or any other such category) not be included at questions unless there is strong justification from other evidence or opinion. Such categories would be likely to increase missing data, reducing data utility. Instead, responders can skip a question if they wish, or be forced to answer a mandatory question with a substantive category when that is critical to census processing and outputs. Decisions need to be made in light of question designs and the voluntary or mandatory status of questions. It should be acknowledged, however, that categories such as “Other” and “In another way” might be used by respondents instead and this would need to be dealt with in processing.

Private response (individual internet access codes or paper forms)

There were mixed views about the need for private response, its potential administration and likely effectiveness as a means of reducing socially desirable responses by trans people living in stealth, reducing the proportion of proxy response and lessening the need for “prefer not to say”.

Views in favour related to:

• trans people with disclosure concerns thus being more able to answer accurately

• it being more appropriate to today’s society that individuals respond for themselves (to all individual questions, not just sex and gender) especially in shared households

There were also views against the method of administration which would involve, including:

• the burden of requesting a private response

• the ability to request a private response being limited to the person completing the household section, which was deemed inappropriate

• the arousal of suspicion this may cause

Serious consideration should be given to improving the method of administration of private response – not limiting the power to request it to one household member – to increase awareness of it within households and empower individuals, without drawing undue attention to themselves. Whether private response by default (for example, providing several individual codes or forms per address at first contact) is pragmatic or feasible might be questionable and could be constrained by technical considerations, costs and the need for one member of a household to be responsible for ensuring all members respond and to respond by proxy when they are unable or are unwilling personally.