Developing sex and gender identity standards: respondent testing findings from our sex question development

About the GSS Harmonisation Team

The Government Statistical Service (GSS) Harmonisation team is based in the National Statistician’s Office, within the Office for National Statistics (ONS). The team:

• is responsible for creating and embedding the UK Government’s “harmonised standards” for all topics
• supports the harmonisation of data across the UK

Our work involves providing bespoke advice and harmonised standards and guidance about how data on different topic areas should be collected and presented to help support statistical producers. The Code of Practice for Statistics recommends that organisations producing official statistics should seek to design new statistics in ways that achieve UK comparability and improve consistency and coherence with related statistics, but does not mandate the use of harmonised standards. Others may also choose to use these standards where they help the consistency, comparability and coherence of data and statistics.

Statistical producers can use the harmonised standards as a starting point in their data collection processes. The standards are designed to promote alignment across organisations, not to enforce identical approaches. Because of this flexibility, they can be adapted to suit specific needs or contexts, including the use of administrative data sources for research. By applying harmonised standards, statistical producers can:

• work in greater alignment with others
• increase the usefulness of their statistics
• ensure that statistics are being used to their full effect for the public good

Harmonised standards allow people to effectively and accurately compare data that have been collected across different datasets, including primary research data and the secondary use of administrative data for research purposes. This means we can more easily understand what those data do, and do not, tell us.

About this report

This report presents the details of our research and respondent testing from our work to develop a harmonised standard for the collection of data on sex. This work aligns with the 2021 report of the independent Inclusive Data Taskforce (IDTF), which recommended regular reviews of harmonised standards for relevant groups and populations, such as those with protected characteristics.

When developing harmonised standards, we use agile methodology and the Respondent Centred Design Framework to ensure that the recommendations we make to statistical producers and data collectors are effective, evidence-based, and fit for purpose. Agile delivery consists of three main phases prior to going live; these are Discovery, Alpha, and Beta.

The Discovery phase is about gathering insights from users and respondents about their needs. User needs refer to the requirements of statistical producers and data users. Respondent needs include usability, acceptability, and question comprehension. We published a first report on our early research findings on sex and gender identity in May 2026, and a report on user needs for harmonised standards on sex and gender identity in October 2026.

The Alpha phase is about using the insights gathered from Discovery research to make evidence-based decisions about question design. Candidate questions are designed, tested through research, refined based on research feedback, and then retested. This iterative improvement process continues until a final question design has been produced that tests well and meets both users’ and respondents’ needs.

This report sets out the research we have undertaken with participants to:

• understand the challenges and differences in interpretation associated with question wording related to the respondent’s sex
• test comprehension and usability of candidate questions for a harmonised sex standard

This report also:

• summarises the findings from our research which led to the published harmonised standard for sex
• discusses the caveats and limitations of our research, and sets out our future research plans

The next phase of work reviewing data collection for sex will align to the Beta phase, where questions can be quantitatively tested at scale.

What we mean by sex

Various ways of asking respondents about their sex have been used within official surveys and data collections over time. This has led to variation in the way that sex is defined, explicitly or implicitly, including:

• variation in terminology (for example, referring to gender instead of sex)
• variations in guidance (for example, for a trans or non-binary respondent on how they should interpret the sex question)

The research findings in this report describe our work towards a harmonised standard on sex data collection. Our testing has prioritised developing questions for two target concepts of sex, aligned with different legislative instruments which have a bearing on an individual’s official records and status:

1. Biological sex (sex at birth) aligned with the definition for the purposes of the Equality Act, 2010,
   and
2. Certificated sex (sex at birth or as modified by a Gender Recognition Certificate) aligned with the legal recognition provided by the Gender Recognition Act, 2004

This reflects the urgent user need for statistical guidance shared with us by official statistics producers, following the UK Supreme Court ruling in April 2025 that sex (for the purposes of the Equality Act) refers to biological sex unmodified by a Gender Recognition Certificate (GRC).

In this report, we use the phrase “sex at birth” rather than “biological sex”. This is because of differences in the way that potential respondents interpreted the meaning of biological sex. The UK Supreme Court defined their use of the term “biological sex” as describing “the sex of a person at birth”. Since this term was less likely to be subject to variations in respondent interpretation, we chose to directly refer to the Supreme Court’s definition as a target concept for data collection.

We use the phrase “certificated sex” to mean a person’s sex at birth, or as modified by a GRC. Individuals may also choose to update the details recorded about them on other official documents, such as their passport and driving license. However, these changes do not require a GRC and are therefore not part of this definition.

We recognise that user needs and the context for official data collection continues to evolve. Therefore, further testing may be needed to update the harmonised standard as circumstances change. We will also continue to research user needs for other topics related to sex, such as variations in sex characteristics (VSC, sometimes referred to as intersex, or differences of sex development).

User needs for a harmonised standard

Our research into user needs for harmonised standards on sex and gender identity identified the following priorities for official statistics producers, data collectors and data users across the GSS:

User need 1

Harmonised standards for sex and gender identity which encourage consistency across collection sources and over time, aiding data comparability.

User need 2

Harmonised standards for sex and gender identity which use clear and distinct terminology and concepts, and are legally compliant.

User need 3

Sex and gender identity questions which meet respondent needs for acceptability, comprehension, inclusivity and information disclosure, and promote trust.

User need 4

Harmonised standards which include guidance for statistical producers on how to implement and administer sex and gender identity questions in data collection, including in administrative data.

User need 5

Harmonised standards which include guidance for data collectors on data handling and presentation.

User need 6

Transparency and practical updates on the harmonised standard development which are regularly communicated.

This report summarises the research we have undertaken to develop questions for data collection which support user need 3 for acceptability, comprehension, and inclusivity. By developing questions which explore distinct target concepts aligned to legislation, this research also supports user need 2. For guidance on using these questions within a harmonised standard (user needs 1, 4 and 5), please refer to the published sex standard.

Nothing in this publication seeks to address user needs for data collection of gender identity. This is an evolving topic, and its inclusion in large-scale data collection is still developing, as described in our previously published research findings. We are prioritising further rounds of discovery research into user needs for data on gender identity, gender reassignment, trans status and trans history before moving into question design and testing on these topics. This research will include the needs of data producers, data analysts, and respondents.

Question development approach

As we learnt from our early discovery research, different concepts and terminology have been associated with sex in surveys and data collection. The way that sex is defined, explicitly or implicitly, has varied across sources and over time. Data quality will be affected if respondents interpret question wording in different subjective ways, as the target concept for data collection will not be effectively understood or captured.

Before generating candidate questions for cognitive and usability testing, we explored respondents’ interpretation of four different styles of questioning on sex. This work provided a background context for the respondent behaviours we would observe when subsequently testing candidate sex questions.

We then conducted two iterations of combined cognitive and usability testing with a new sample of respondents. To do this, we present candidate questions within a prototype online questionnaire and have respondents engage with the questions and verbally ‘think through’ how they read and respond. This work is carried out in line with Respondent Centred Design Principles. We design for an online mode of presenting the question as our first mode of data collection because if a design can be easily used and understood in a self-complete mode, then it stands a high chance of being easily used and understood in an interviewer-led mode.

Different conditions were explored for the candidate questions, for example with and without respondent guidance. During testing, the order in which questions and conditions were presented to respondents was rotated, to mitigate order effects.

We then iterated to a second set of design options based on respondent behaviour and what we learned about the reasons for that behaviour through their verbalised thought processes when engaging with the first set of candidate questions. This new iteration of candidate questions were taken through a further round of cognitive and usability testing with another sample of respondents.

Two iterations of Alpha testing provides a ‘silver’ approach to cognitive testing. That means that there is still a level of risk associated with the final standard, although significantly less than a design which relied on ‘at-desk’ design or only a single round of testing. Our research identified some groups with whom a further iteration of testing would be desirable in order to meet a ‘gold’ standard approach to RCD. This is discussed in a following section on the “Limitations and next steps” of the research.

Sampling for qualitative research

Qualitative research is a crucial tool for developing an understanding of the rich variation of interpretations and complex issues that affect the way different groups of respondents engage with question designs. When recruiting for qualitative studies, the principles applied to sampling for quantitative studies are not transferrable. Qualitative methods involve purposive samples rather than random or representative samples. This means that research is carried out with small and targeted pools of participants because they can provide information directly relevant to the purpose of the research. In our development of a harmonised standard for sex we recruited a greater proportion of transgender participants than would have been included in our research sample included without purposive recruitment. Census data in England and Wales and census data in Scotland suggests between 0.4% and 0.5% of the population are transgender, however this group includes those respondents most likely to interpret questions in distinct and varied ways compared to non-transgender respondents. Our research therefore targeted a larger trans population sample of around one third the total research sample in order to understand where these differences may occur.

Language testing

We conducted 28 interviews with a purposive sample of UK respondents recruited using an external recruitment agency. We conducted a one-to-one interview with each participant using a semi-structured topic guide to gather qualitative data, which was then analysed through coding and thematic analysis. The interviews sought to explore respondent conceptualisations of sex language in a data collection context. Four different data collection questions were shared, and respondents were asked to comment on what they understood the question was asking. The order in which questions were presented to respondents was rotated, to mitigate order effects.

The questions we used for background language testing, which included questions and guidance drawn from the independent review of statistics and research on sex and gender, Census 2021 in England and Wales, and Scotland’s Census 2022, were:

Cognitive and usability testing

We carried out two rounds of iterative “cogability” (cognitive and usability) testing during our development of the sex harmonised standard. This testing allows us ensure that questions, guidance, response options and any other respondent materials:

• are understood as intended
• are usable
• meet respondent needs
• are not burdensome for respondents

We recruited a total respondent pool of 91 participants using an external recruitment agency. 50 participants were interviewed during the first round and 41 during the second. Our primary research sample comprised of 81 participants who had the following characteristics:

Sex

Age

UK location

Ethnic group

Trans status or history

We did not seek to specify the definition of sex used in our screening questionnaire and instead used the common question stem from our previous language testing: “what is your sex?”. This means the sample characteristics shown in the tables do not account for the potential for transgender participants to have varying interpretations on how to answer the question about sex. When considering only those 55 individuals who reported that they did not have a trans status or history (and therefore were unlikely to respond differently depending on their interpretation of the question intent), 32 people indicated that their sex was female and 23 indicated that their sex was male.

In addition to the primary research sample, we recruited two purposive ‘top-up’ sample groups in our second iteration. These comprised 5 participants who answered “prefer not to say” to our screening question on trans status or history, and 5 participants whose main language was one of Urdu, Bengali, or Turkish. This second group was selected based on the primary languages spoken in areas of England and Wales where ONS research identified the greatest uncertainty in gender identity estimates from Census 2021. Additional research with this population was targeted as a first test for issues in language comprehension, although this should be supplemented at a later stage with follow-up interviews among low-proficiency English speakers.

For each participant, we conducted a one-to-one combined cognitive interview and usability test by presenting candidate questions within a prototype online questionnaire. Each participant was presented with candidate questions for both biological sex (sex at birth) aligned with the definition for the purposes of the Equality Act, and certificated sex (sex at birth or as modified by a GRC). Different respondent guidance conditions were tested for each question, and the order in which questions were presented to respondents was rotated, to mitigate order effects. Analysis of the qualitative data captured in this stage focused particularly on participant behaviours and themes from their verbalised accounts of question interaction (including comprehension and response selection) as they completed each question condition. We did not use feedback on how the participant believed others might interpret the question (referred to as ‘advocacy feedback’) in our analysis in this phase. This was because the primary purpose of testing was to validate whether the question design was understood, and whether it produced the desired behaviour, in practice.

We did include a small number of additional qualitative questions with the top-up sample of respondents for whom English was not their primary language. These questions sought to explore potential comprehension issues with other speakers of their language. The additional data collected through this route did not surface any particular challenges. However, this would be considered ‘advocacy feedback’ rather than feedback from the group of interest, so we cannot rule out that such issues may be found. We will use the data and feedback associated with this top-up sample of participants to assist in recruitment and research design in future work with individuals who have lower English language proficiency.

Findings

Question for biological sex (sex at birth) aligned with the definition for the purposes of the Equality Act 2010

The following question design was the most successful method for collecting sex at birth:

During testing we found that most non-transgender participants skimmed over all variations of questions on sex, including both question wording and guidance. They relied instead on recognition of the response options to identify the question purpose, and determine their response. This finding was observed among non-transgender respondents in all our test conditions, but was particularly consistent among participants in our second iteration top-up sample for whom English was not their primary language. Across our testing conditions, we found the primary source for difference in response behaviour – and therefore data quality – was among transgender respondents. In data collection it is important that the respondent understands the question and answers it in a way that aligns with the user need for the data. It is also important that respondents are willing to provide their data to the researcher through the data collection instrument. Respondents who find a question unacceptable to answer, or who do not recognise a response option through which they can identify themselves are more likely to drop out of research altogether, creating gaps in the data.

In preliminary language testing we identified particular challenges with questioning on sex at birth for transgender respondents. In general, transgender participants reported understanding what was expected in a question about sex, and the differences between sex and gender identity. However, this was more complicated when a respondent had engaged in a process to change their sex, including but not limited to seeking a GRC. Guidance was therefore necessary for this group to provide appropriate instruction for participants on how to interpret a sex question. Transgender participants frequently interpreted a specific reference to sex at birth as relating to birth documentation, and this association supported them in identifying how they were being asked to assign themselves. However, the direction to respond based on sex at birth was uncomfortable for this group. Feedback during our language testing phase indicated this group would be more likely to disengage with data collection without a clear and agreed rationale for the purpose of gathering such data. This creates a risk to data quality through data gaps, where a demographic group sharing particular characteristics are more likely to be missing in official statistics due to survey drop-out or non-participation.

Our first iteration of a candidate question to collect sex at birth addressed these issues by including guidance on the specific target concept. We trialled this in different conditions by including the term ‘at birth’ into the question wording, and by using additional supplementary respondent guidance. For supplementary guidance, we drew on the automatic links which transgender participants made to birth documentation, and referred to ‘the sex recorded on your birth certificate’. Lastly, we used the past tense and asked ‘what was’ the respondent’s sex at birth, in order to reduce survey drop-out behaviour and address potential confusion for respondents who had undertaken processes to change their sex.

This first iteration question performed well with both transgender and non-transgender respondents. For both populations, specifying the term ‘sex at birth’ within the question wording meant that the purpose of the question was widely understood. While some non-transgender respondents questioned why ‘at birth’ needed to be clarified in the question, their understanding of the question and the ease of identifying an appropriate response option for themselves was not affected. Indeed, many non-transgender respondents did not notice the specifics of the question wording, reporting that the question felt ‘familiar’ and ‘a standard question’ to see in a questionnaire.

The past tense framing of the question specifically helped transgender respondents to understand how to answer, and partially improved the likelihood that they would answer. However, this effect was much more notable when supplementary respondent guidance referring to the sex recorded on a birth certificate was presented alongside the question. This was particularly valuable for non-binary respondents who were more likely to disengage with a survey question which only asked for sex at birth without such guidance. Among this group in particular we found that referring to what was recorded on a document added clarity and made the question less emotive. While including this supplementary guidance had the potential to add burden for non-transgender respondents by increasing the amount of reading required, we found that their likelihood to skim over the question actually meant that including additional guidance referencing a birth certificate was a positive enabler. This was because the inclusion of a reference to a familiar document made them feel supported in their first impression of the question intent. While there were some individuals for whom the past tense of the question created potential confusion, this was not found when the supplementary guidance was provided.

Our second iteration of a candidate question to collect sex at birth therefore maintained the question stem which had performed well in iteration one, and introduced variants of supplementary question guidance (sex registered at your birth, and sex recorded on your birth certificate). We also updated the guidance to emphasise the past tense when referring to the individual’s birth certificate by asking about “the sex that was recorded on your birth certificate”. This was to ensure clarity for transgender participants with a GRC who might have subsequently requested a new birth certificate.

The consistent use of past tense throughout the question and guidance performed well in terms of producing an easy to answer question. Transgender participants appreciated the intent of the question as relating to a point in time. They again reported that this, and the framing of additional guidance as focused on official documents, was important for making them feel more comfortable with the question and more likely to continue to engage. Non-transgender participants frequently skipped the guidance entirely, finding it unnecessary to instruct them in selecting a response. They reported that the question stem was clear on its own, felt familiar and straightforward, and were sure of their response. This ease of responding was consistently true even among a smaller number of non-transgender respondents, who noticed that the question stem differed from other sex questions they had answered.

Finally, we found the reference to a birth certificate in the supplementary guidance was more easily interpreted by respondents than the phrasing ‘registered at your birth’. Several participants, both transgender and non-transgender reported that the phrase ‘registered at your birth’ required them to pause and decipher, or consider what was expected of them. In resolving this confusion, participants proactively began referring to their birth certificates as the solution to the question of how their birth was registered.

Question for certificated sex (sex at birth or as modified by a Gender Recognition Certificate)

The following question design was the most successful method for collecting sex inclusive of GRCs:

During testing we found that most non-transgender participants skimmed over all variations of the question on sex, including both question wording and guidance. They relied instead on recognition of the response options to identify the question purpose and determine their response. However, this pattern was disrupted when a reference to ‘Gender Recognition Certificate (GRC)’ was made near the beginning of the respondent guidance, as this terminology was frequently unexpected or unfamiliar to them. During language testing we also found a wide range of reactions to guidance referencing a GRC from transgender participants. For some participants this created confusion over whether the question was seeking data on sex or on gender. For some, it was seen as an inclusive way of allowing a response in line with acquired sex. Others reported that they felt it did not represent inclusivity as they were still directed against responding in a way which matched their current identity. We also noted some negative feedback to guidance which began “if you are considering how to answer…”. Some participants interpreted this phrasing to imply they were confused over their identity, rather than over the intent of the question.

Our first iteration of a candidate question to collect sex inclusive of a GRC used a simple question stem (“What is your sex?”) with two guidance conditions. In one condition we prompted respondents with “If you have a Gender Recognition Certificate (GRC), you can use the sex recorded on this legal document”. In the other condition we included a second, final sentence “Otherwise, use the sex recorded on your birth certificate”.

For non-transgender participants, the question stem was recognisable and easy to answer. We saw a continuation of the pattern in which many of this population skimmed through the question, relying on the familiarity of the response options to progress through the survey. However, for other respondents, the guidance added confusion and was negative in terms of their ability to identify a response. This was because, for some respondents, the phrasing in condition 1 (“if you have…”) was interpreted to mean that individuals who did not meet the guidance condition should not answer the question at all. Including an ‘otherwise’ instruction in condition 2 improved this, however participants were primed by their initial confusion over the term ‘Gender Recognition Certificate’ and – while able to answer the question – found it burdensome to interpret the entire guidance. The inclusion of the term “legal document” was also found to add respondent burden as respondents felt this meant they would have to answer more carefully.

Transgender respondents also responded more negatively to this question and its associated guidance. This included confusion over how they should answer if they did not have a GRC in condition 1, but also a particular concern that this question (while referring to their sex in the present tense) did not reflect their identity today. We found this heightened still further as an issue for non-binary respondents, either because:

• it reinforced for them that a GRC would not enable them to legally recognise their identity
• they were unfamiliar with the term and anticipated that the reference to ‘gender’ in guidance should offer a greater range of response options in line with their gender identity rather than their sex

Our second iteration of a candidate question for sex inclusive of a GRC sought to address the confusion over who should answer the question which was created by the “if you….” phrase; and to provide a reference to a more commonly known document to make the whole guidance easier to interpret. We were also interested to understand whether definition guidance (using the descriptive phrasing “by sex we mean…”) which had performed well in the sex at birth testing, was more effective or less effective than directive guidance (using the instructive phrasing “you should use…”). We therefore maintained the previous question stem, and used two guidance conditions:

1. “You should use the sex recorded on your birth certificate or Gender Recognition Certificate (GRC)”
2. “By sex we mean the sex recorded on your birth certificate or Gender Recognition Certificate”.

In this second round of testing, we saw a significantly improved response, in terms of:

• engagement with the question
• response accuracy to the question intent
• respondent burden

Among non-transgender participants, the question was considered to be comfortable to answer. As the “if you have…” wording had been removed, this group understood that it was intended for them to answer. Re-ordering the references so that the familiar “birth certificate” was included first in the guidance made the question easy for them to answer, despite not knowing what a GRC was. They did not feel the need to seek more information about this as they were readily able to focus on the part of the guidance applicable to them. We also saw a return to the previous ‘skimming’ behaviour, as several participants could not identify any difference between the guidance in condition 1 or 2.

For transgender participants, the question was easy to answer, but less comfortable. It was perceived to be inclusive by recognising sex recorded on a GRC, however for some respondents this also reminded them of the difficulties of obtaining a GRC, and the associated emotions of being asked to record their sex in a way that did not reflect their current identity. This mental and emotional processing added burden to transgender participants engaging with the questionnaire. However, they were able to correctly interpret the intention of the question, although some participants who held a GRC reported having to pause and consider whether to record the sex from their GRC or their birth certificate.

The ‘definition’ guidance was perceived more positively in this set of tests than the ‘instructive’ guidance. For transgender respondents, this is because it allowed for a definition of sex to be established which was localised to the intention of the specific question, increasing their willingness to engage. In general, respondents reported that the guidance would improve their accuracy in responding.

Response options for asking about sex

For both target concepts of sex, we provided binary response options of “Female” and “Male”. This aligns with the response options used in the most recent UK censuses, including the use of alphabetical ordering. These response options were considered appropriate and acceptable in testing with non-transgender respondents, who in many cases used their familiarity with these terms to guide their engagement with the questionnaire over the specifics of the question wording. For non-binary and transgender respondents, the limitation to a binary response was discussed as a source of emotional tension when completing the questions, however this was reduced in effect when the questions included guidance referring to an individual’s birth certificate or GRC. In these conditions, the response options were considered more appropriate as they aligned with the only possible sex options which could be recorded on these official documents. This enabled transgender respondents to subjectively differentiate between the answer being recorded on the questionnaire, and their experience of their own gender identity.

Context for asking about sex

Across testing for both sex at birth and certificated sex, many respondents reported that they would want to know the justification for collecting data on sex. This was frequently raised by transgender respondents, but also by non-transgender respondents who expressed wanting to be sure there was a legitimate reason for being asked this question. Respondents said that their expectations on the relevance of sex data collection would vary by the identity of the data collector – for example, having greater trust in a large statistical survey than in data collection for commercial purposes.

For transgender respondents, the relevance of asking about sex, and a particular target concept of sex, could change their behaviour in answering a survey question. If there was no clear justification, they reported that they might answer the same question differently or decide not to continue completing the questionnaire, which would result in issues of non-response bias for researchers. These rounds of question testing cannot quantify this risk for several reasons. Firstly, it can be expected that participants willing to be recruited for research on survey design are more likely to be positive about data collection and surveys. Secondly, due to the qualitative nature of question testing research, the findings on specific behaviours such as survey drop-out cannot be generalised to produce a quantitative impact assessment. However, our findings on the discomfort transgender participants associated with disclosing information on sex which does not align with their preferred gender identity is consistent with similar findings reported in question development testing for sex and gender identity topics by National Records Scotland (PDF, 1,464KB). We therefore recommend that researchers ensure information is available to respondents setting out both the:

• purpose of data collection for sex within their study
• user need underlying decisions made about the target concept of sex being sought

Several transgender respondents indicated a preference for answering questions based on their gender identity rather than sex. We are also aware that some GSS data collectors have sought to make data collection more inclusive by enabling survey respondents to answer a question about their gender identity in addition to a question about their sex. The GSS Harmonisation team are conducting further research into user and respondent needs for data and questions on gender identity, trans status, and gender reassignment. The team expect to publish a harmonised standard for gender identity in Summer 2027. As part of this work, further testing will be carried out to understand how respondents interact with a questionnaire which contains both sex and gender identity questions, including the impact of question order and signposting. As a preliminary step to help plan this research, we added supplementary testing conditions to the second iteration of our sex question testing to display signposting text that indicates a question on gender identity would follow. Findings from these supplementary conditions will be used alongside other qualitative research with respondents from a range of backgrounds and characteristics to understand what guidance is needed for statistical collectors considering this approach.

Limitations and next steps

The research presented here includes two rounds of iterative question testing with respondents from a range of backgrounds and perspectives. This means that our design is aligned to the ‘silver’ level of the Respondent Centred Design framework.

To extend our research to align to ‘gold’ level, we will consider additional question testing among specific respondent groups. These represent particular populations who may require variation in questions to support their understanding of the question target, such as individuals:

• for whom English is not just a second language, but who have a low English language proficiency
• with learning difficulties
• with variations in sex characteristics (VSC, sometimes referred to as intersex, or differences of sex development), who may have corrected their birth certificate or be considering doing so
• who have migrated to the UK from countries without comparable birth registration and certification processes
• who are responding on behalf of someone else (also known as ‘proxy’ respondents)

Testing for the sex standard has so far focused on development for an online self-completion mode, applicable for survey or administrative data collection in English. This is in line with the RCD framework which recommends taking an ‘optimode’ approach to design, where respondent materials and questionnaires are tailored for each mode and explored through combined cognitive and usability testing. When designing for multiple modes, we always design the online mode first and then move on to develop the other modes. We do this because if a design can be easily used and understood in a self-complete mode, then it stands a high chance of being easily used and understood in an interviewer-led mode. We want to expand this testing in future to explore how the sex standard may be tailored for other modes of data collection (for example, paper, telephone, or face-to-face), and to data collection in Welsh. This may result in the inclusion of alternate variants of the sex question specific to the requirements of different modes or languages. This is an approach we have used previously in other harmonised standards, for example presenting alternate variants of questions which use show-cards for face-to-face interviewing.

The qualitative respondent testing described in this report is part of our Alpha phase. We are recommending the question designs as templates for wider use, as we consider the evidence supporting them to be substantial enough to move into larger-scale review. The next phase of work reviewing data collection for sex would align to the Beta phase, where questions can be quantitatively tested at scale. We are seeking feedback from data collectors who are looking to adopt these questions, to monitor their performance in different survey contexts.