The Census 2021 question on sex is “What is your sex?” with response options of “Female” and “Male”.  Sex can be defined in different ways.  This paper considers five concepts of sex that have been identified. All are valid concepts for data to be collected on, this evaluation is in the context of the Census.  These are sex as: registered at birth, recorded on birth certificate, recorded on legal/official documents, living / presenting, and self-identified.

Four key uses of sex data from Census 2021, plus wide-ranging needs for aggregate data tables, were identified during the consultation and engagement.  These are for:

1. population estimates and projections and household projections;
2. 2021 Census operational needs, including coverage adjustment process;
3. equality monitoring; and
4. analysis and linkage of ONS Longitudinal Study and microdata sample files.

There are separate needs for data on gender identity, and a separate question “Is the gender you identify with the same as your sex registered at birth?” with response options “Yes” and “No, write in gender identity” is included later in the form. The key Census 2021 sex data user needs have been evaluated in relation to relevance, and comparability and cohesion.

Relevance

Each identified data need has been assessed against current use, and potential use, in relation to existing methodologies, time series and admin data sources. By considering the sources and methods of collecting data, and uses of the data, it is concluded that the most evidenced needs are for sex recorded on legal/official documents and sex as living / perceived.

Comparability and coherence

User consultation responses focussed on a need for continuity over time and across the UK.  We would aim for data collected to be comparable with all UK censuses and other social surveys, coherent with the wording of the 2021 census question (which asks ‘what is your sex’) and the other uses of the word “sex” in the census questionnaire.

By considering the approach across UK census offices and on UK surveys, it is concluded that the most evidenced need in relation to this dimension of quality is sex recorded on legal/official documents, with similar, but lower, levels of need for sex as living / perceived.

The scale of the impact of collecting any specific concept of sex cannot be determined as multiple factors are at play.  Both coverage and response rates are important as low response rates could mean a longer field period or extended estimation processes, and larger coverage adjustment could lead to wider confidence intervals around the estimates.  However, 6 per cent of the total population of England and Wales (3.8 million people) did not respond to the 2011 census and were adjusted for during data processing. In comparison, the trans population is small, estimated to be between 0.3 and 0.75% of the UK population.

The impact on Census 2021 has been assessed in relation to accuracy and reliability.  We evaluate different concepts of sex in relation to impact on overall census response, sex question response, and individuals right to privacy.

Accuracy and reliability

This evaluation considers the 2021 Census context, for example that it is a mandatory household survey administrated by online or paper self-response by one or more household members. Two strands of additional research were completed using best practice qualitative research methods. These included in-depth interviews with trans people, traditional women’s groups and the general population and a review of the evidence base, to help understand what sex concept will:

• maximise Census responses from the trans population / the cisgender population
• maximise sex question responses from the trans population / the cisgender population
• ensure any perceived invasion of privacy is justified by the data need

Sex is a key variable to ensure a genuine census record which, if not completed, could lead to a record being removed during processing.

The highest negative impact is anticipated if sex registered at birth was collected due to reduction in census response. The data need is not considered to be strong enough to justify collecting this.

The least impact is anticipated if sex as living / presenting is collected. However, the adjacent concept of sex recorded on legal/official documents is also anticipated to have low impact.

The evaluation of the evidence presented supports the collection of sex as registered on legal/official documents in Census 2021 for England and Wales.

Guidance will seek to minimize barriers to completion of the census and the sex question.  To do this it needs to balance the following aims:

• direct people in how to answer;
• be considerate of people that have identities other than male or female or have variations of sex characteristics / are intersex;
• recognize that some respondents may have legal/official documents that state male and legal/official documents that state female; and
• use terminology that is acceptable to the whole population.

To ensure we collect the characteristics that are needed by data users, we conducted the 2021 Census topic consultation in 2015.  This was a topic level consultation and received 1,095 responses of whom 249 provided a response on their data need for demographic information, which included the sex question.

The consultation responses highlighted the user need for continuity and a high response rate.  None of the consultation respondents defined the concept of sex to be collected in their initial response.

The key needs identified were for population estimates and projections and household projections, 2021 Census operational needs, including coverage adjustment process, equality monitoring, and analysis of ONS Longitudinal Study and microdata sample files.

In some cases, they stated that gender rather than sex should be collected.

In most cases, the request was not specific enough to identify a specific concept because cultural as well as biological factors impact the proposed research.  Similarly, in the previous Beyond 2011 consultation (2014), and the subsequent Census 2021 Output Content Design consultation (2018) sex is proven as a key variable, but no specific concept has been requested.

Since the 2015 consultation, ONS has continued to engage with a range of stakeholders regarding the sex and gender identity questions. Much of this is reported on in the publications listed in Annex B.  This annex also provides further information on engagement with women’s groups that has been ongoing since September 2017.

In December 2019, 80 social statisticians and users of population level data signed a letter to the directors of UK census authorities with concerns about the sex question guidance.  This outlined concerns with the concept of sex to be collected around understanding:

• sex-based discrimination
• different health and socio-economic outcomes between men and women

ONS met with some of the signatories of the letter in January 2020 and subsequently held a roundtable with a wide range of academics who use data on sex in June 2020.

The roundtable showed that there were a range of data needs amongst the expert academic community and no consensus was reached.  Therefore, the ONS undertook further research, including follow-up meetings with groups of academics with different views to understand further their data requirements.

Each identified data need has been assessed against current use, and potential use, in relation to the following questions:  What would we aim to collect:

• considering the existing methodologies?
• if we did not already have a time series?
• for comparability with administrative data sources in the future?

The relevant ONS methodology teams have been consulted for their assessment of the impact of the sex question on the population estimates and projections and housing projections.  Their response is provided in full in Annex C.  In summary these outputs use a range of information, including:

• Census estimates: This is interpreted as self-identified sex both in 2011 and before, although in 2001 guidance to that effect was only available on request, and prior to that no guidance was provided.
• Birth registrations: This is the only key source that is explicitly sex registered at birth and is collected directly from the register of births.
• Death registrations: The sex is provided by the person registering the death, so it may not be sex registered at birth.
• International Passenger Survey 3: The interviewer typically records what sex they believe the respondent to be.4
• GP Patient Register: Here applicants are invited to tick whether they are “Male” or “Female”, with no indication of the definition to be used (or even whether it is sex or gender which is being asked for).
• Annual Population Survey. There is no guidance, but we expect the interviewer typically records what they believe the respondent to be.
• Ministry of Justice prisoner data. For the purposes of the projections, all people in men’s prisons were counted as male, and all those in women’s prisons as female in the data supplied[5].  Location within the prison estate is decided by an initial local transgender case board or escalated to other boards if required.  All decisions are made on a case by case basis.

Overall, sex in the source data is largely undefined and can be respondent recorded or recorded by a proxy such as an interviewer. The dominant source is the 2011 Census data and, apart from birth registrations, the other sources may also include people with sex recorded differently from their sex registered at birth.

In addition:

• the projections need the best possible census coverage of the population of England and Wales, with undercount minimised and the aim to meet the national completion target of at least 94%.

ONS plan to develop a revised back series of population estimates for the period 2011 to 2021, which in turn will inform projections. For this, methodological consistency over time, and at either end of the revised series, would be optimal.

Table 2: Assessment of population and household estimates and projections needs

The relevant ONS methodology teams have been consulted for their assessment of the impact of the sex question on the 2021 Census estimates.  Their response is provided in full in Annex C.  In summary the need is for:

• Non-response and multi-ticking: High levels of either of these would negatively impact the statistical imputation processes as sex is a key predictor variable.  There would be a greater risk of bias in imputation of a number of variables.  The ‘Remove False Persons’ process uses sex as a key variable, so high levels of missing information could lead to more records being removed, although the method could be adjusted to use a different variable instead.
• Collection of the same concept in Census 2021 and Census Coverage Survey: Matching the Census to the Census Coverage Survey is key for coverage estimation and imputation.  The sex variable is used mainly to indicate a non-match when they differ, although matches can be made automatically when sex is different or missing. Our assessment is that there will be no matching bias caused by collecting different sex concepts.

The census coverage survey is interviewer led and the sex question is asked.  If queried the interviewer would access the main census question guidance.  However, experience from ONS social surveys show that in a face-to-face interview the concept is rarely questioned – respondents answer as they wish.  Three instances of interviewers being asked how to answer were raised in the LFS in the last 3 years.  In the face-to-face Census coverage survey respondents respond as they wish.  It is likely that sex as living / presenting will be collected as no reason to provide sensitive personal data is provided.

Table 3: Assessment of 2021 Census operational needs, including coverage adjustment process

Sex is a protected characteristic, and so data on sex informs Equality Impact Assessments.  The Equality Act (2010)[6] states:

“In relation to the protected characteristic of sex—

• (a) a reference to a person who has a particular protected characteristic is a reference to a man or to a woman;
• (b) a reference to persons who share a protected characteristic is a reference to persons of the same sex.”

The status of being a man or a woman is not further defined within the act, however the Equality and Human Rights Commission (EHRC) state, in their statement on sex and gender reassignment legal protections and language that a person’s sex is what is on a person’s current birth certificate.

The EHRC page on sex discrimination (February 2020) further explains that:

“The Equality Act 2010 says you must not be discriminated against because:

• you are (or are not) a particular sex
• someone thinks you are the opposite sex (this is known as discrimination by perception)
• you are connected to someone of a particular sex (this is known as discrimination by association)

In the Equality Act, sex can mean either male or female, or a group of people like men or boys, or women or girls.”

Therefore, a member of the trans population should be protected from discrimination due to their:

• sex registered at birth (due to protection against direct discrimination), and
• sex as living / presenting (due to protection against discrimination by perception)

As gender reassignment is a separate protected characteristic, discrimination on those grounds are not considered here[7].

Table 4: Assessment of equality monitoring needs

Detailed analysis of record level census data is conducted by a wide range of data users following linkage processes carried out by ONS.  There are two main record level sources, and lists of research projects using each of these are available publicly:

• ONS Longitudinal Study for England and Wales [8]
• Census microdata files

The ONS Longitudinal Study is a 1% sample of England and Wales and includes linked census and life events data from 1971 to 2017. It was originally set up to enable detailed health research.  However, it has since been used for analysis on all Census topics.  These include studies looking at differences between the male and female population.

The ONS Longitudinal Study includes a range of data sources, and retains the variables as recorded in each one.  This includes:

• Sex on census records
• Sex as registered at birth (for those born in the England and Wales)
• Sex on life events records, for example cancer registrations

The Census microdata files only contain 2011 Census data, which is interpreted as self-identified sex.

Table 5: Assessment of Analysis of ONS Longitudinal Study and microdata sample files needs

There is no single concept of sex required across the four primary data needs.  All concepts of sex are needed to some extent.Many other uses of census data at an aggregate level are quoted, for example in gaining wider understanding of those giving care, education, health, socio-economic status, deprivation, migration, commuting habits, demand for local bus services, domestic abuse patterns, the gender pay gap.  In each case insufficient detail was provided to assess the precise need, but inferences can be made for different topics for different concepts.  Therefore, table 5 also reflects these needs.

The most evidenced needs are for sex recorded on legal/official documents and sex as living / perceived. The evidence for sex recorded on legal/official documents is stronger when looking forward to increased use of administrative data in the future, whereas the evidence for sex as living / perceived is stronger in relation to existing methodologies.

Consultation responses focussed on a need for continuity over time and across the UK.  In reviewing this, assumptions must be made about how people answer in different circumstances, and whether how people answer is aligned with the guidance.  This means that in many cases more than one concept has been recommended.

This is an area where internationally the approach being taken is evolving, and examples of all approaches can be found.  As such, this evaluation does not consider comparability beyond the UK.

Table 6: Assessment of comparability and coherence needs

Evaluation of relevance showed that the most evidenced needs were for sex recorded on legal/official documents and sex as living / perceived.  Consideration of comparability and coherence shows no need for sex as registered at birth.

The most evidenced need in relation to this dimension of quality is sex recorded on legal/official documents, with similar but lower levels of need for sex as living / perceived.

What sex concept will:

• maximise Census responses from the trans population / the cisgender population?
• maximise sex question responses from the trans population / the cisgender population?
• ensure any perceived invasion of privacy is justified by the data need?

Sex question response can impact overall response as sex is a key variable in the ‘remove false persons’ process.  Therefore, if this variable is not completed, a record is more likely to be removed from the dataset during processing.

Each concept will be given a red, amber or green status against each evaluation question.  Where a group is not discussed in the evaluation, this means that there was no evidence from testing of the concept under discussion leading to higher non-response for that group.

Testing showed (2020:6) that the potential impacts on overall census response and response to the sex question were largely the same across groups.  Therefore, the evaluation considers both together and, in table 7 the term “respond” refers to both responding to the census as a whole and to responding to the sex question.

Outcomes were assigned against the following evaluation criteria:

• Green: No evidence of negative impact on overall response or sex question response across all groups either at the individual level or at a societal level
• Amber: Evidence of negative impact on overall response or sex question response for subsets of society either at the individual level or at a societal level
• Red: Evidence of negative impact on overall response or sex question response across a wide range of groups at the individual level and at a societal level

Table 7: Evaluation of what sex concept will maximise Census responses

An individual’s right to respect for their private life is set out in Article 8 of the European Convention on Human Rights (1950), which is incorporated into UK law by the Human Rights Act (1998). Article 8 states:

Everyone has the right to respect for his private and family life, his home and his correspondence.

There shall be no interference by a public authority with the exercise of this right except such as is in accordance with the law and is necessary in a democratic society in the interests of national security, public safety or the economic well-being of the country, for the prevention of disorder or crime, for the protection of health or morals, or for the protection of the rights and freedoms of others.

This does not prevent interference by a public authority with a person’s right to respect for their private life, but it requires certain conditions to be met for that interference to permitted by the Human Rights Act. Firstly, the interference needs to be in accordance with the law, and secondly it needs to be both a necessary and proportionate means of achieving one of the stated permitted purposes. Further information can be obtained from the Equality and Human Rights Commission.

Outcomes were assigned against the following evaluation criteria:

• Green: Data need sufficient to justify collection of data, AND trans respondents would not be forced to reveal their gender history.
• Amber: Data need insufficient to justify collection of data, OR trans respondents would be forced to reveal their gender history.
• Red: Data need insufficient to justify collection of data, AND trans respondents would be forced to reveal their gender history.

Table 9: Evaluation of what sex concept would ensure any perceived invasion of privacy is justified by the data need

The least impact is anticipated if sex as living / presenting is collected.  However, the adjacent concept of sex recorded on legal/official documents is also anticipated to have low impact.The highest negative impact is anticipated if sex registered at birth was collected due to reduction in census response.  The data need is not considered to be strong enough to justify collecting this.

Question context

The phrasing and display of the sex question as presented on the EQ and the paper questionnaire imply that sex, not gender should be provided.  However, the precise concept of sex to be provided is not stated.

• The term “sex” is used
• There are only two response options
• There is reference to a question on gender identity to follow.

The question does not have hard validation, as only questions that are required to build the form have this applied to allow respondents to more easily navigate around the form.

Image 1: Census 2021 sex question, online self-completion for those aged 16 and over

On the electronic questionnaire the ‘Help’ link in the bottom banner brings up the “Online questions help” index page in a new tab.

Paper questionnaire respondents are provided with a link to www.census.gov.uk/help on the front page of the paper questionnaire.  From there a respondent must select the link for “Paper questions help” to reach the index page. This is a separate, but very similar page to the online equivalent.

The headings at the top of the page are anchor links that take the user further down the same page to the links for individual guidance. These then take the user to a new page with guidance on the question of interest.

The person must move between tabs, or close the new tab, to return to the EQ.  There will be no button on screen allowing the person to quickly return to EQ.

Image 2: Draft Census 2021 online questions help index page

Information on stakeholder engagement and testing has been provided over time as the project has developed through a series of update reports from the point of the 2021 Census topic consultation onwards.  Census legislation is included in the below listing in italics to show key points of Parliamentary engagement

Sex and gender identity are only the focus of research in the tests listed in the question development report.  However, as sex is a key demographic variable, it was included in all tests and feedback incorporated into the sex question decision making as required.

For reference, the paper forms are also available: Paper questionnaires for England and Wales.

These publications evidence our engagement with a range of stakeholders.  Further evidence, of more recent engagement, is provided in the next section

1. May 2016, Consultation response Demography (1.6 MB pdf)
2. August 2016, Gender identity workshop Gender identity workshop
3. January 2017, we published a Gender identity update
4. June 2017, Gender identity update event
5. September 2017, Qualitative research on gender identity: phase 1 summary report,
6. December 2017, 2021 Census topic research report
7. July 2017, 2021 Census Roadshow slides Gender identity
8. December 2018, 2021 Census topic research update  research and testing of questions and topics for the 2021 Census.
9. December 2018, government White Paper Help Shape Our Future: The 2021 Census of Population and Housing in England and Wales (PDF 886KB)
10. February 2019, how “sex” and “gender” are used in the context of the UN Sustainable Development Goals (SDGs)
11. June 2019, The Office for National Statistics (ONS) and Statistics Canada (STC), as part of the UNECE Bureau of the Conference of European Statisticians, collaborated to undertake an In-depth review of measuring gender identity
12. September 2019, Measuring sex in the UK censuses: Harmonisation update
13. September 2019, Guidance for questions on sex, gender identity and sexual orientation for the 2019 Census Rehearsal for the 2021 Census
14. October 2019,  Census (Return Particulars and Removal of Penalties) Act 2019
15. March 2020: Summary of testing for Census 2021 (March 2020)
16. May 2020, Census (England and Wales) Order 2020
17. June 2020, Census (England) Regulations 2020, Census (Wales) Regulations 2020 and Rheoliadau’r Cyfrifiad (Cymru) 2020
18. June 2020, Exploring existing data on gender identity and sexual orientation
19. June 2020: Sex and gender identity question development for Census 2021
20. June 2020: Paper questionnaires for England and Wales
21. 19 October 2020, Demography question development for Census 2021

The summary of testing, first published in March 2020, has recently been updated with four further tests undertaken in 2020.  As such, it contains further details on all the tests referenced in this report.  These are summarised in table 10.

Table 10: Summary of testing used in the evaluation of the sex question concept

Following the consultation in 2015, ONS continued to engage with those stakeholders who had consented to further contact regarding the sex and gender identity questions.

Our September 2017 research report on initial work on the sex and gender identity led to increased levels of engagement, in particular with users of data on the female population.  ONS proactively identified women’s groups to engage with who had undertaken any research on sex, used any statistics which depended on sex or who appeared to have an interest in the collection of sex data (and possibly gender identity data too).

In April 2018 a stakeholder event ‘Understanding Data Collection on Sex and Gender Identity’ was held in London.  Stakeholders were consulted about the work that ONS were doing on gender identity and why it was being conducted. They were asked for their views on the sex and gender identity question options for the 2021 Census.

Representatives from the following organisations were invited and representatives from the organisations listed in bold attended the event.

• Women’s Institute
• Women’s Budget Group
• Equality Now
• The Equality Trust
• Women’s Aid
• Gender and Development Network
• Equality and Diversity Forum
• Equality Challenge Unit
• Demos
• The Women’s Resource Centre
• Sex and Gender Ethics Society
• Fawcett Society

ONS confirmed that there will be a mandatory question that will collect male and female information.  There was a consensus from most stakeholders that although they understood the user need for gender identity data, they would not want this to impact the data currently collected by ONS on sex as this data is used for various purposes in their organisation.

Some cross-government working / topic groups were regularly attended:

• Harmonisation topic group
• Gender Identity Cross-Government topic group
• Scottish National Statisticians Sex/Gender Working Group

Other organisations were met with individually, these organisations we met with in 2019, when the Census 2021 questions on sex and gender had been agreed, are listed below.  Those we met with more than once on this topic are marked with a star (*)

• a:gender
• Centre for Equalities
• Civil Service HR
• Central local information partnership (CLIP) population sub-group
• Care Quality Commission *
• Department for Culture, Media and Sport
• Department for Education *
• Equality and Human Rights Commission *
• Family Resources Survey Conference
• FemQuant
• Government Digital Service
• Government Equalities Office *
• Intersex UK
• Ipsos MORI
• LGBT Foundation Stonewall
• Market Research Society
• Medopad
• Ministry of Justice
• National LGB&T Partnership
• Natsal
• NHS Digital
• NHS England *
• Northern Ireland Statistics Research Agency *
• National Records for Scotland *
• National Statisticians Data Ethics Advisory Committee
• Office for Standards in Education
• ONS social surveys
• Stonewall *
• Universities and Colleges Admissions Service (UCAS)
• Understanding Society
• United Nations Economic Commission for Europe
• Welsh Government

In December 2019, 80 social statisticians and users of population level data sent a letter to the directors of UK census authorities with concerns about the sex question.  This outlined concerns with the concept of sex to be collected around understanding:

• sex-based discrimination
• different health and socio-economic outcomes between men and women

These concerns are addressed in the discussion of user needs in section 2.  In addition, concerns were raised regarding:

• Size of the population for whom concepts change response (addressed in section 3.1)
• Impact on other data collection exercises (addressed in section 2.3)
• Reference to those with intersex conditions (addressed in section 5)

ONS met with some of the signatories of the letter in January 2020.  It was agreed that ONS would then hold a Roundtable with a wide range of academics who use data on sex.  Due to COVID, this was delayed until June 2020.  The roundtable event included people from the following organisations:

• com
• LGBT Foundation
• Government Equalities Office
• Equality and Human Rights Commission
• NHS England
• Anglia Ruskin University
• Stonewall
• UCL
• University of Edinburgh
• ESRC
• Imperial College London
• London School of Economics
• King’s College London
• University of Kent
• University of Manchester
• University of Essex
• University of Edinburgh

Since the 2015 consultation, ONS has continued to engage with a range of stakeholders regarding the sex and gender identity questions. Much of this is reported on in the publications listed above.

The roundtable showed there were a range of data needs amongst the expert academic community and no consensus was reached.  Therefore, the ONS undertook further research, including follow-up meetings with groups of academics with different views to understand further their data requirements:

• July 2020, meeting with data users supporting collecting data on sex registered at birth
• August 2020, meeting with data users supporting continuing to collect data on self-identified sex

ONS did not present any research at any of these meetings, instead allowed the stakeholders to present their views.  None of the meetings yielded further evidence of data needs for specific concepts of sex in Census 2021.

Centre for Ageing and Demography response:

The population projections (and derived household projections) use the previous Census as their base. This is currently the 2011 Census. The 2011 Census sex question guidance indicated that people could self-identify. Exact wording was:

“Select the answer which you identify yourself as. You can select either “male” or “female”, whichever you believe is correct, irrespective of the details recorded on your birth certificate. You do not need to have a Gender Recognition Certificate.”

In practice this means that the projections are fundamentally based on respondent-identified sex, noting that:

• Most respondents will not have looked at the guidance, so (as with earlier Censuses) will have self-identified anyway. The proportion of the population who looked at the sex question guidance was 0.003% (a bit less than 1 in 30 thousand people).
• A respondent’s self-identified sex may or may not be their sex as registered at birth. For the cis population, it would presumably be so. However, for the trans population this would be split.

Projections also take account of population changes over time through births, deaths and migration:

• Births are projected through a combination of:

(i) sex ratios based on sex at birth, taken from birth registrations – this is the only key source that is explicitly sex at birth

(ii) female fertility rates. The latter are derived with the number of births as the numerator and the number of females as the denominator. The number of females in that denominator comes from the Census-based population estimates. This means that:

• It will include some trans people (born male) as well as biological females.
• It will also include biological females who are infertile.
• It will exclude those who were registered as female at birth, but recorded their sex as male in the Census [This is because they will be counted in the male population].

• Deaths are projected through death rates. These are derived from:

• Number of deaths, by sex, taken from death registrations. The sex is provided by the person registering the death, so it may not be sex at birth. For those who lived in a gender different from their sex at birth, they may well have been recorded in their lived gender (indeed, we cannot assume that the person registering the death will even know the deceased’s sex at birth). There is no linkage made with birth records.
• Population estimates. As noted previously, these are primarily derived from Census and so self-identified.
  • Migration is derived from a range of admin and survey sources. These are also liable to be sex as defined by the person completing the form.
    • The main international migration source to date has been the International Passenger Survey (IPS), where the interviewer would typically record what they believed the respondent to be. Clearly in most cases this will be the same as sex at birth, but people who have transitioned to living in a new gender may well have been recorded in that gender.
    • The main internal migration source is the GP Patient Register, where applicants are simply invited to tick whether they are “Male” or “Female”, with no indication of the definition to be used (or even whether it is sex or gender which is being asked for). So for many transgender respondents, it may well be lived gender, different from the sex registered at birth.

The household projections also use 2001 Census data. This is also self-defined (there was no written guidance for the 2001 Census although we have evidence that ONS advised individual contacts that they could self-identify). They also use the Annual Population Survey, which also isn’t necessarily sex at birth. In addition (like the population projections) they use prisoner data from the Ministry of Justice (MoJ).

• For the purposes of the projections, all people in men’s prisons were counted as male, and all those in female prisons as female – i.e. this is assumed to reflect “lived sex”, not necessarily sex at birth.
• According to published MoJ statistics, in 2019 there were 163 transgender prisoners in England and Wales (out of an overall prison population of over 80,000, so a rate of around 0.2%). Of these, 34 were in women’s prisons and 119 in men’s prisons.

In summary, sex in the projections is predominantly respondent-identified. The dominant source is the 2011 Census (self-identified) and, with the exception of birth registrations, the other sources are also likely to include people recorded differently from their sex at birth.

In that context, indicating that people to self-identify in the 2021 Census will be most consistent. Note too that PSD is likely to develop a revised back series of population estimates for the period 2011 to 2021, which in turn will inform projections. On that front too, having self-identification for 2021 will offer most methodological consistency over time, and at either end of the revised series.

In addition, the projections need the best possible coverage of the population of England and Wales, with undercount minimised and aiming to meet the national completion target of at least 94%. Therefore, it could be argued that any guidance which is likely to deter people from answering is disadvantageous.

Census and Population Statistics Hub response:

Edit and Imputation

Missingness

The sex question had very low missingness in 2011 (less than 0.5%). It was a key matching variable used to find nearest neighbours in the donor imputation process because it has been identified as a key predictor variable for other census characteristics.

We are currently investigating scenarios around high item non-response caused by potential controversy around the sex, gender identity and sexual orientation questions. These include higher non-response of the sex question by people or groups unhappy with the published guidance. Potential impacts could be: a smaller donor pool for imputation leading to a greater risk of bias in imputation which could affect a number of other census variables, loss of a key predictor variable for a significant number of records, again leading to a greater risk of bias in imputation for other census variables, and increased time needed to run imputation to resolve missingness leading to more time pressure on the rest of the imputation process and perhaps subsequent processes.

High item non-response from the trans population itself may slightly increase the overall non-response rate for sex. This is unlikely to cause issues with sex distributions in isolation after imputation, but imputation of sex will not take into account the gender identity question, so there is a small risk around the maintenance of the multivariate relationship between sex, gender identity and other variables after imputation for the trans population specifically. No research is planned to investigate this before the census due to the anticipated small magnitude of the impact.

Multi-ticks (where both sexes are selected on the paper questionnaire) could also lead to an inflated missingness rate as sex would be set to missing for these cases, so will also be factored into scenario testing. This could apply to cisgender (where both sexes selected as a protest), trans and intersex populations.

Inconsistencies

Following changes to legislation, there are no hard edit rules around marital/civil partner status and sex so any inconsistencies in the way trans people may answer the question would not have an effect on edit rule failures for this rule.

There are hard edit rules surrounding age of mothers compared with their children – a mother cannot be more than 65 years older than her child. The way transwomen answer the sex question could have a small effect on edit rule failures for this rule but this is likely to affect only a tiny proportion of records so it has been considered unlikely to have any significant measurable effects on imputation.

Census Linkage

In 2011, there was less than 0.5% and 1.8% missingness on the sex variable in the census and census coverage survey respectively. The percentage of matches made where sex was the same, missing or different was 98.2%, 1.4% and 0.4% respectively. In 2021, we again expect sex to be generally well recorded and accurate. Whilst records matching on sex is not a strong indicator of a match, records not matching on sex is a strong indicator of a non-match. Therefore, sex is used as a matching variable in the majority of our deterministic matchkeys and as a probabilistic matching variable. However, it is still possible for matches to be made automatically when sex is different or missing. Cases that are not matched automatically by either deterministic or probabilistic methods, will be reviewed clerically and could therefore be matched if the records do refer to the same person, despite there being a difference or missingness in the sex variable. We therefore do not anticipate that there will be a bias in matching caused by advice that is, or is not, given regarding the sex question. Lastly, if in 2021 we find that there is a much higher than expected rate of missingness in the sex field then we could react to that by tweaking the matchkeys and/or the parameters for the probabilistic matching.

Census Coverage Estimation

Estimation is based on the collection of data from the Census Coverage Survey. The requirement is that the same concept is captured in both census and CCS. If there happen to be any differences in response between the Census and CCS (this can be due to a number of reasons such as collection, capture or imputation), the census value is used for estimation. Estimation would only be impacted if the guidance affected overall response or matching rates significantly.

Census Coverage adjustment

This imputes records into the census database so that the overall database is consistent with the census estimates from the estimation process.  No particular concept is required as long as the same concept is collected in both the census and the census coverage survey, which is identical to the estimation requirement.

Statistical disclosure control

There would be no impact on the SDC methods used. Sex is not a targeted variable in record swapping, and it would have no impact on the rules used as part of the flexible table builder.

Removing false persons

Sex is a key variable in this process, however if levels of missingness were significantly greater than anticipated the methodology could be adjusted to use a different variable with higher item level response.

id=”Foot1″ 1 A very small number of people with VSC/intersex may be classified as “indeterminate” in their first weeks of life.  These are reclassified as either male or female unless the infant died before this took place.

2 This concept refers to UK issues legal/official documents.

3 Data collection for the International Passenger Survey halted in March 2020 due to coronavirus (COVID19) pandemic and restarted in December 2020.

4 In future, international migration statistics will be determined by administrative sources. While research on this is ongoing, it is clear that the sex record in these may also not match sex at birth (for example, work visa applications require a passport as evidence but, dependent on where the migrant has come from, we know this may not be their sex at birth).

5 Peterborough Prison is the only prison within England and Wales with both male and female residents.

6 Further information on the act is available in the gov.uk guidance

7 Further guidance is available on gov.uk at Equality Act 2010: guidance

8 There are separate longitudinal studies in Scotland and Northern Ireland.

9 Further information on the 2011 Census quality and methodology is available in the 2011 Census Quality and Methodology Information report.