Sex and the Of ﬁ ce for National Statistics: A Case Study in Policy Capture

On 17 March 2021, just four days before the England and Wales census date, the Of ﬁ ce for National Statistics (ONS) conceded in a judicial review case. The case concerned the meaning of the sex question in the census, and the guidance to accompany this question. How did the ONS end up in court defending its guidance on a question which most people would deem self-explanatory? This article explains the legislative and political context around gender recognition, and the consequences for data collection. Drawing on my own experiences, I explore the politicisation of the de ﬁ nition of sex, and argue that the ONS gave undue weight to lobby groups with a particular viewpoint on sex and gender identity.


Introduction
'When I use a word', Humpty Dumpty said, in rather a scornful tone, 'it means just what I choose it to mean-neither more nor less.' 'The question is', said Alice, 'whether you can make words mean so many different things.' 'The question is', said Humpty Dumpty, 'which is to be master-that's all.' 1 THE UK CENSUS has gathered information on sex since its inception in 1801, and rates of non-response to the sex question have remained very low. 2 Yet, the Office for National Statistics (ONS) sought to redefine sex to encompass gender identity for the purposes of the 2021 England and Wales census. With just days to spare before the census date, the ONS was forced to concede that it did not have the authority to redefine sex in this way, following a preliminary judicial review hearing where their arguments were given short shrift. This outcome raises the question: how and why did the ONS find itself in court defending a position which appeared to lack both coherence and legal foundations?
The 2021 census included the compulsory question 'What is your sex?' with the standard binary response options 'Male' and 'Female'. For the first time, there was an additional voluntary question on gender identity: 'Is the gender you identify with the same as your sex registered at birth?' (Yes/No), with an open text response field for those choosing 'No' to write in their gender identity however they chose. The aim of the gender identity question was to allow people with trans and non-binary or other gender identities to express this. These two distinct questions should have made it possible to maintain the principle that sex and gender identity are distinct characteristics, and to capture them separately. However, draft ONS guidance issued in 2019 proposed that respondents could answer the sex question in terms of their subjective gender identity, undermining the scientific value of both responses. This guidance was influenced by lobbying in the context of proposed reforms to the Gender Recognition Act (GRA 2004). The proposed reforms would have allowed individuals to change their legal sex by a simple self-identification (gender self-id). The proposals were dropped by the government in 2020, but the self-id lobby remains influential. This article traces how this lobby gained influence in ONS, how attempts by social researchers 1 1 to ensure that statistical practices adhered to objective definitions were resisted, and how the battle ended up in the courts.
Sex is a fundamental demographic and explanatory variable, yet in recent years, sex has been overwritten by gender identity across a range of data collection exercises. I have written elsewhere about the reasons that social scientists and others are concerned about this. 3 One reason for the focus of concern on the census in particular is the tendency for other data collection exercises to treat the census as the gold standard, meaning that the wording of its questions and guidance set an important precedent. This paper focusses on the England and Wales census, though the discussion is equally relevant to the Northern Ireland and Scotland censuses, directed by the Northern Ireland Statistics and Research Agency (NISRA) and National Records of Scotland (NRS) respectively. Indeed, the push to overwrite sex with gender identity in law, policy and data collection is international, with the same tactics of policy capture and suppression of debate. 4 This article investigates the politicisation of the basic demographic category 'sex', and the way in which lobbyists have attempted to change the meaning of the word sex in order to undermine the collection of clear data on sex. Political campaigning by adherents of gender self-id has been overtly directed at securing legal change. However, genderidentity lobbyists have also campaigned behind the scenes for changes in policy and practice, including in relation to data collection. 5 The process whereby lobbyists representing a particular political perspective develop close relationships with governmental organisations, allowing them to co-produce policy without due consideration for competing interests, has been termed 'policy capture'. 6 In postmodernist style, the goal of the gender-identity lobby has been to change reality by redefining the meaning of words. Such Humpty Dumptyism, while prevalent in certain academic niches, jars rather glaringly with the precepts of good questionnaire design.
A notable feature of campaigning in favour of gender self-identification is the silencing of discussion. 7 Women asserting that sex matters in this context have suffered penalties, including the loss of their livelihoods, for example in the case of Maya Forstater. 8 This chilling climate has stifled discussion on data collection as it has on other questions relating to sex.
The census produces data for quantitative social research; specifically, in the ONS's standard formulation, to provide information to guide the allocation of public services. It is appropriate that the ONS consults with those who are going to use the data when developing questionnaires; indeed working with 'user groups' is a longstanding feature of ONS practice, as with other social research organisations. As I will show, ONS deviated from this usual practice in this instance, bringing in stakeholders who were not data users, and who expressed opposition to the systematic collection of data on sex.
Accurate data on sex is not a trivial matter. It is sometimes assumed that replacing information on sex with gender identity will have only a small effect on data accuracy. However, this relies on assuming low and stable numbers of incongruent cases. The numbers of people who consider that their gender identity differs from their sex is unknown, but there is evidence of rising numbers among youth, particularly girls. 9  very wide range of estimates for the proportion of youth whose gender identity is incongruent with their sex. 10 It is impossible to predict how these trends will change over time. Clear information on sex is vital for understanding such change. For example, we cannot examine why girls are increasingly expressing trans and non-binary identities if collecting data on sex is taboo. But inaccurate data on sex does not only undermine the value of data on gender identity: small numbers of misallocated cases can have a large effect on research findings in any sub-group analysis where one sex is dominant. One obvious example is violent crime, where a small number of males identifying as female can greatly skew the sex ratio (this is even more stark in the case of sexual offences, where 97 per cent of perpetrators are male). 11 The fact that trans people are not randomly distributed in population subgroups is also relevant. For example, if trans women are disproportionately likely to work in certain highly male-dominated employment sectors such as IT, disallowing data based on sex may create an illusion of progress towards gender parity in these fields. 12 Conversely, if females who identify as trans or non-binary are more likely to take maths and physics at university, data based on gender identity alone could make it appear that women's representation in these fields is going into reverse. The direction of potential biases cannot be assessed without the collection of accurate data on both sex and gender identity.
I begin by outlining the legislative framework and campaign for legislative change regarding gender recognition, which provides necessary context to understand the ONS's actions. Next, I give a brief overview of the history of the sex question guidance. I present an account of the process through which ONS developed and revised its guidance from 2019 to 2021. I provide unique insight by combining first-hand experience with information obtained via freedom of information and subject access requests. Where I cite correspondence, this is available on request. I describe the arguments put forward at the judicial review hearing. I conclude by noting the importance of transparency and trustworthiness in statistics, and the dangers of policy capture by lobby groups, and suggesting some remedial actions.

Legislative and political context
The Gender Recognition Act (GRA 2004) was an obscure piece of legislation until an attempt to radically broaden its reach brought it to public attention. In 2017, the Conservative government opened a consultation into reform of the GRA. The proposed reforms would have removed medical oversight and gatekeeping, allowing people to simply self-identify their preferred legal sex without the requirement of a diagnosis of gender dysphoria. New grassroots women's organisations including Woman's Place UK (WPUK) and Fair play for Women (FPFW) sprung up to demand discussion of the proposed changes, which would have had sweeping implications for issues affecting women and girls, including sports, single-sex spaces and data collection. In 2020, the Minister for Women and Equalities, Liz Truss, announced that gender selfidentification would not be passed into law. Nevertheless, lobby groups promoting gender self-id, and arguing for the removal of the sexbased exemptions in the 2010 Equality Act (EA2010), most prominent among them Stonewall, had already entrenched their position within a range of institutions, including government departments. The ONS appeared enthusiastic about the proposed changes in 2017, referring to 'gathering momentum for change', and claiming that 'Within today's society the traditional view of gender as a binary classification, male or female, is changing.' 13 Until 2005, a person's sex as recognised in law (hereafter 'legal sex') was determined by 10 Murray Blackburn Mackenzie (MBM), 'International evidence and the risks of reframing the sex question in the census', 30 November 2020; https://murrayblackburnmackenzie.org/2020/11/30/ international-evidence-and-the-risks-of-reframing-thesex-question-in-the-census/ (accessed 15 June 2021). 11  their biological sex, which is observed and registered at birth, but this changed in the UK when the GRA came into force that year. The GRA allows people with a diagnosis of gender dysphoria to change their legal sex, providing them with a Gender Recognition Certificate (GRC), and entitling them to have a new birth certificate issued with the preferred sex recorded. The GRC is an example of a 'legal fiction', and it is acknowledged that it has no bearing on one's actual biological sex. 14 A clear example of an exemption in law, where biological sex always overrides a GRC, applies to the descent of peerages. The slogan 'Trans women are women and trans men are men' does not apply if one's father is a duke. Other exemptions, including for sports and single-sex services, can be invoked where they represent 'a proportionate means of achieving a legitimate aim'. The EA2010 treats sex and genderreassignment as separate protected characteristics, and also underlines the sex-based exemptions to the GRA.
The GRA was the UK government's response to a 2002 ruling by the European Court of Human Rights that the government was in violation of Article 8 (regarding the right to privacy) and Article 12 (regarding the right to marriage) of the European Convention on Human Rights. Although civil partnerships were brought in in 2005 to grant legal standing to same-sex unions, the GRA was introduced because the government at the time preferred to allow a small number of people to change legal sex, rather than allow same-sex marriage. (Legislation permitting same-sex marriage ultimately came into force in 2014.) The right to marriage dominated the discussion of the bill. 15 To date, it is estimated that approximately 6,000 people have a GRC.
The legislation gives GRC holders the right to 'privacy', in the following sense: 'The Gender Recognition Act 2004 safeguards the privacy of transsexual people by defining information in relation to the gender recognition process as protected information. Anyone who acquires that information in an official capacity may be breaking the law if they disclosed it without your consent.' 16 This right to privacy is not absolute, and was introduced in the context of a discussion focussing on personal relationships and marriage, with clear reference to a small group of transexual people whose natal sex might potentially be unknown to their associates, and for whom it would be undesirable for this information to become public. It is not obvious how such a right could clash with any anonymised data collection exercise. Yet, the right to privacy has been interpreted by the ONS and others to have implications for data collection. Data collection is never mentioned in the Hansard record of the discussion of the bill, and it seems highly unlikely that legislators anticipated such repercussions.

Historical overview of sex in the England and Wales census
The census has included information on sex since its inception in 1801. The Census Act (1920) provides the legal foundation for the census, making it lawful for the government to carry out a census, and specifying the matters in respect of which particulars may be required, including sex.
Self-id guidance was first put in the public domain in the 2011 census. The guidance was online, while data collection was paper-first, which meant that the guidance largely went unnoticed. The guidance appears to have been implemented without any consultation with data users or any assessment of its effects on data quality. Until 2017, the ONS made changes under the radar, without public attention, but this changed when the ONS proposed to make the sex question in the 2021 census 14 A. Asteriti and R. Bull, 'Gender self-declaration and women's rights: how self identification undermines women's rights and will lead to an increase in harms: a reply to Alex Sharpe, "Will gender selfdeclaration undermine women's rights and lead to an increase in harms?"', Modern Law Review, vol. 83, no. 3, 2020, p. 539. 15  voluntary. 17 The ONS made the recommendation following focus group consultations with trans people, some of whom claimed the sex question was 'irrelevant, unacceptable and intrusive'. 18 Prominent academics objected to the idea of dropping the mandatory question on sex. WPUK also objected to ONS's proposals and launched a parliamentary petition opposing the proposal. 19 The ONS backtracked in the face of this public opposition. Nevertheless, it had not dropped the idea that there was something problematic about asking respondents for their sex. In 2019, the ONS released self-id guidance for the census dress rehearsal, as follows: What is your sex? Select either 'Female' or 'Male'. If you are one or more of non-binary, transgender, have variations of sex characteristics, sometimes also known as intersex, the answer you give can be different from what is on your birth certificate. If you're not sure how to answer, use the sex registered on your official documents, such as passport or driving licence, or whichever answer best describes your sex. A later question gives the option to tell us if your gender is different from your sex registered at birth, and, if different, to record your gender.
The 2019 rehearsal guidance clearly indicated that some respondents may answer the sex question with a response other than their biological sex or legal sex. The ONS eventually revised its guidance in February 2021 following concerns raised by quantitative social scientists. 20 It now read as follows: Please select either 'Female' or 'Male'. If you are considering how to answer, use the sex recorded on one of your legal documents such as birth certificate, Gender Recognition Certificate, or passport. A later question gives the option to tell us if your gender is different from your sex registered at birth, and, if different, to record your gender identity.
This revised guidance reflected some movement towards acknowledging the concerns of the scientific community. However, it fudged the target of the question by adding a document which does not necessarily record legal sex, namely passports, while the phrasing 'such as' introduces further ambiguity. The judicial review decision led to the words 'such as' and 'or passport' being struck out. The final guidance therefore reflected legal sex, as follows: 'If you are considering how to answer, use the sex recorded on your birth certificate or Gender Recognition Certificate.' Silencing discussion on sex and data I coordinated a letter from eighty quantitative social scientists raising concerns with ONS and the other UK census authorities regarding the guidance to accompany the sex question, published in late 2019. 21 ONS had not consulted quantitative social scientists on the guidance, and the scientific community was largely unaware of the proposed change in the meaning of the sex variable. The letter instigated a protracted and strained process of attempting to represent the views of the signatories to ONS.
I was initially naïve enough to imagine that denial of the material reality and importance of sex would not be enforced in the rational world of social statistics. I was astonished, therefore, to be no-platformed from a research methods seminar on the measurement of sex and gender identity. I have discussed this incident elsewhere, but have previously refrained from discussing the role of ONS given that negotiations regarding the census were ongoing, and I hoped that an outcome driven by 17 scientific considerations was attainable. 22 It now seems important to document that ONS appeared to be complicit in cancelling an opportunity to engage with mainstream views on the measurement of sex.
The event in question was jointly organised by NatCen (National Centre for Social Research) and City University. Apart from myself, there were four confirmed speakers when I spoke to the organiser after I was initially invited: one from NatCen (Nancy Kelly, subsequently CEO of Stonewall), one from its sister organisation ScotCen, and two from ONS, including Helena Rosiecka (Head of 2021 Census Question and Questionnaire Design, and author of the report justifying ONS methodology, which we will come to later). On 23 December 2019, I received an email to panel members stating: 'I regret to inform you that we have decided to cancel the survey methods seminar that was scheduled for 21 January.' I was already aware that Guy Goodwin (NatCen CEO) had instructed NatCen staff not to sign the letter of concern to the census authorities.
I suspected that the event may have been cancelled in order to de-platform me, and therefore requested a meeting with Guy Goodwin. I also put in a subject access request for all NatCen emails discussing me by name. I met with Goodwin and another member of the NatCen senior team on 8 January. Goodwin has close ties to ONS, as he came to NatCen in 2016 from the ONS, where he was Head of Profession for Statistics. Goodwin acknowledged that the event had indeed been cancelled to avoid platforming me. He stated that, while not the main reason, a 'subsidiary factor' in cancelling the event was the 'mood music' at the ONS. Goodwin stated that people at the ONS (unnamed) were unhappy about the public letter of concern; they had apparently questioned my motives, and claimed that I was not going through proper channels. I was shocked to hear of this hostility from the ONS. Goodwin also told me that one or more of the speakers had threatened to pull out of the seminar if I was allowed to speak.
The emails retrieved via subject access request showed that a staff LGBT group had asked that I should not be platformed. The emails also supported the suggestion of ONS knowledge of and engagement in this no-platforming. An email set up a phone call between the organiser of the seminar and the speakers (excluding myself) to discuss 'the inclusion of Alice Sullivan' and the possibility that I should not take part or that the event should be cancelled. These speakers, as mentioned, included two ONS staff. Another NatCen email discussing how to go about no-platforming me, referred to 'the not fighting with ONS aspect'.
Attempting to engage with ONS I requested a meeting with the National Statistician, Ian Diamond, by email on 18 December 2019. I was granted a meeting with Iain Bell, the Deputy National Statistician, on 22 January 2020, and attended with a colleague. During this meeting, Bell stated that decisions on the guidance would be finalised during the period July to September 2020. In the event, the guidance was not finalised until February 2021, with mere weeks to go before the census date, for reasons which were never explained. Bell suggested a round table meeting with academics and 'other stake-holders'. I stressed the importance of hearing the views of quantitative social scientists in particular, as they had not yet been consulted.

The round table
The round table was held on 24 June 2020. Six quantitative social scientists who had signed the letter of concern to the census authorities were invited. ONS also invited six academics who all supported the self-identification view and were opposed to asking people's sex, objectively defined. These academics, with one exception, could not be described as experts in survey design or data analysis. Their departmental affiliations included theology and religious studies, material sciences, law, and gender studies.
The ONS later cited a lack of consensus in this meeting, implying that experts were divided on the question of what the sex question is intended to measure. But this lack of consensus was artificially generated. One could question why, if there is a lack of consensus, ONS was unable to identify six pro self-id academics with quantitative social science 22 Ibid.
6 AL I C E S U L L I V A N expertise, and instead had to resort to inviting non-data users.
Alongside the academics, six representatives of policy groups and government bodies were in attendance. Five put forward pro self-id views: Stonewall's Head of Policy, the LGBT Policy Director at the Government Equalities Office, and delegates from the LGBT Foundation, the Equality and Human Rights Commission, and NHS England. It was notable that the government representatives advocated the gender self-id perspective, and none commented on the implications of sex-based data collection for women's rights. Dr Nicola Williams from FPFW was the sole policy presence advocating for accurate sex-based data as a necessity for equalities monitoring.
Bell concluded the meeting by saying that, since consensus had not been reached on the guidance for the census sex question, ONS would organise further research and testing. The research proposals would be shared with the group along with the results.

Follow-up meeting for data users
Following this meeting, I wrote to Iain Bell, pointing out that there had been very little time for data users to get their points across in such a large and diverse group, and requesting a meeting for data users only. Following a reminder, we were granted a forty-five minute appointment in July. We discussed the research which ONS proposed to do, and tried to establish what ONS believed the 'target' of the sex question to be-that is, what exactly this question is meant to measure. ONS was (and subsequently remained) unwilling to give a clear answer to this point. We questioned how any methodological assessment could be made of the impact of the guidance wording on data accuracy without taking a view on what information the sex question is intended to collect.
We were subsequently told that the planned research had been cancelled. Yet, months later, it seems the research was revived. We were never shown the results, despite repeated requests.

Eleventh hour changes
Sir Ian Diamond was interviewed on the BBC Radio 4 Today programme on 22 January 2021. He indicated that the sex question in the 2021 census would be based on legal sex, as recorded on a person's birth certificate. It was surprising to hear such an unequivocal statement, since there had been no communication regarding a decision being reached on the guidance.
After contacting the Chief Statistician for clarification, I was granted a meeting with Diamond and Bell on 28 January, which I attended with a colleague. This represented the first direct contact I had with Diamond since first contacting him in December 2019. At this meeting, it became clear that ONS was considering a form of wording which would not in fact specify legal sex, as stated on the Today programme, but sex on any documents. When challenged, Diamond acknowledged the vagueness of this wording, and indicated that he would be happy to tighten the wording so that it actually referred to legal sex, that is, the sex on one's birth certificate. We noted that the signatories would prefer the guidance to refer to sex registered at birth, but that a legal sex variable would be better than a self-id variable, and that loose wording should be avoided. We asked to see the research which ONS had carried out and were told they did not know when it would be published, but they would share it with us on the basis that it would be kept within the group of eighty academics. A note circulated by ONS after the meeting confirmed this.
We were then summoned urgently to a meeting on 5 February, where we were told by Ian Diamond that clear guidance indicating legal sex was not in fact an option. Diamond indicated that, following the previous meeting, Bell had told him that this was not possible. The justification given was 'legal advice' from a government department. Diamond's public statement that the 2021 census guidance would indicate legal sex suggests two possibilities: either that he was inadequately briefed on this issue, or that he preferred not to be transparent about a contentious issue on national radio.
The arguments that I and other academics were presented with throughout this process shifted and twisted under challenge. Crucially, we were unable to pin down what ONS deemed the target of the sex question to bethat is, what were they trying to capture with this item in the census? We were presented with inconsistent views on the legal position. Research evidence was referred to but never shared.
Shifting arguments on the law The Equality and Human Rights Commission (EHRC) advised the Scottish government in 2019 that collecting data on respondents' sex would be a potential violation of their human rights, particularly their right to privacy and dignity under Article 8 of the European Convention on Human Rights and under the Human Rights Act 1998. 23 ONS appears to have been influenced by the same line of argument, which was contested in a legal opinion commissioned by WPUK. Aidan O'Neill QC's opinion states that collecting data on sex is lawful, so long as it is a proportionate means of achieving a legitimate aim. 24 I was initially told by the Deputy National Statistician, Iain Bell, that it would be illegal to ask about natal sex. Subsequently, when I sent Bell a link to Aidan O'Neill's opinion for WPUK, he responded that the ONS approach is research-based, and therefore legal advice did not form any part of its thinking, only to return to legal considerations being deemed paramount at our final meeting. On being challenged about the nature of the legal advice, Bell has admitted that it is legal to ask a person's sex if this is deemed to be a proportionate means of achieving a legitimate goal. At one point, Bell pointed to avoiding the risk of judicial review as a reason for settling on legal sex 'as being birth certificate or gender recognition certificate'. Yet, in fact it transpired that such wording had not been settled on at all and subsequently, Bell claimed that government legal advice presented an insurmountable obstacle to asking either natal or legal sex.
Throughout this process, there was no indication that the legal advice referred to was new or had changed. The ONS would not share its legal advice, or a summary of it, in any form. Bell claimed that legal advice is privileged information which cannot be shared (this is incorrect: an individual can choose to share legal advice they have been given, though they are not obliged to do so). This lack of transparency made it impossible to know what the true legal parameters were on the question guidance-and, therefore, to what extent methodological considerations were relevant.

The elusive research evidence
Additional question testing was proposed at the June 2020 round table. Bell stated his intention to share the research design, and later the findings, with participants. We were later told that the planned research had been dropped.
Yet, months later (February 2021), we were told the research had been carried out after all (in November 2020), and we were promised that we would be shown the results. In fact, the results, though referenced in ONS's response to the Methodological Assurance Review Panel (MARP), have never been shared with the participants in the round table or with the signatories of the letter of concern, as had been promised. A short overview of the research was later put into the public domain. 25 A fuller account of the research was only produced in response to a FOI request. 26 The methodology 23 Sex and Gender Working Group, meeting, 23 September 2019, Equality and Human Rights Commission submission; https://www.gov.scot/ binaries/content/documents/govscot/ publications/minutes/2019/09/sex-and-genderin-data-working-group-meeting-september-2019/ documents/ehrc-submission-on-collecting-andpresenting-data-on-sex-and-gender/ehrcsubmission-on-collecting-and-presenting-data-onsex-and-gender/govscot%3Adocument/EHRC% 2Bsubmission%2Bon%2Bcollecting%2Band% 2Bpresenting%2Bdata%2Bon%2Bsex%2Band% 2Bgender..pdf (accessed 15 June 2021). 24 WPUK, 'EHRC misrepresents the law on collecting sex data', 11 December 2020; https:// womansplaceuk.org/2020/12/11/ehrcmisrepresents-the-law-on-collecting-sex-data/ (accessed 15 June 2021). 25  is not compelling. Interviews were conducted with fifty-two people, and the sampling strategy is frankly odd, targeting individuals with selected political views, including 'trans allies' and members of groups described as 'traditional women's groups', which included a 'sex worker led collective'. 27 Women's groups which had raised questions about the guidance were not selected, yet it was concluded that the ONS's preferred wording would minimise the risk of a campaign response. This research did not in fact test the question wording proposed by the ONS in February 2021, making it difficult to see how it could be used to support this wording.

The role of statistical oversight bodies
Myself and a number of other academics contacted both the Office for Statistics Regulation (OSR) and MARP to ensure that they had some oversight of the situation. The purpose of the OSR is to enhance public confidence in the trustworthiness, quality and value of statistics. MARP had the remit of providing external, independent assurance and guidance on the statistical methodology underpinning the 2021 census.
In response, and as part of the ongoing OSR assessment of the census authorities' compliance with the code of practice, Ed Humpherson of OSR wrote to the ONS in September 2020 raising concerns about both the transparency of the research process, and the need to consider the impact on data quality for subgroups of the population (my emphasis in quote): 'ONS should seek to address outstanding concerns raised by users within its further question testing and research on the guidance on the sex question. ONS should share this research in a transparent and open way … The assessment team thinks it essential for ONS to consider the concerns raised by users … and consider the impact of data quality on small sub-groups of the population.' 28 ONS complied with neither of these requests.
In October 2020, Sir Bernard Silverman, Chair of MARP, asked to see ONS's working: 'In line with the code of practice's requirement for "transparent judgements about definitions and methods" it is necessary for ONS to come to a clear view of what definition it wishes respondents to use. The Panel is expecting an opportunity to comment on the guidance in the light of that view'. 29 Silverman also set out additional expectations regarding user engagement and the documentation of this engagement, referring explicitly to the group of experts who had written to him raising concerns: 'As well as more informal contact with users and experts, they will presumably be given written feedback addressing issues they have raised, at an appropriate level of detail and rigour.' This kind of feedback was never produced for the expert signatory group. Silverman added: 'On less contentious topics, the Panel has been given general assurances about stakeholder engagement and has been happy to take these on trust', but requested a more detailed account of ONS's consultation with data users in this instance. Silverman's intervention led to the publication of the MARP paper, described below.
basics'. 31 Nevertheless, the final version was signed off by MARP the following month.
The ONS paper begins with the claim that it will unveil the 'concept of sex' to be collected from the sex question. Yet, the author seems not to understand the difference between a concept and its operationalisation. We are told that there are five concepts of sex: 1) registered at birth; 2) recorded on birth certificate; 3) recorded on legal/official documents; 4) living/presenting; 5) self-identified (often referred to as gender/ gender identity). Yet, of these, only the first two can be said to correspond to a concept of sex. Sex registered at birth reflects biological sex, and sex on a birth certificate (which may be changed only if one has a GRC) reflects legal sex. The other 'concepts' do not reflect sex, but seek to confound sex with gender identity.
The report concludes 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.' It is in no way obvious how this conclusion is reached. Throughout the report, claims are made without any clear reference to relevant supporting evidence. It seems remarkable that a report of such poor quality was signed off by MARP. 32 The disingenuousness of the report is apparent in the section on user engagement, which Silverman had specifically requested: ONS … 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 followup meetings with groups of academics with different views to understand further their data requirements.
As I have noted, several of the academics at the round table did not 'use data on sex', because they did not use data at all, nor could they be said to have 'data requirements'. Rather, it appears that they were invited simply to manufacture the lack of consensus which ONS required in order to side-line the views of the relevant scientific community. 33

ONS's relationship with Stonewall
Stonewall has historically been a hugely important organisation in advancing gay rights in the UK. However, since 2014, its primary focus has been on promoting gender self-id. Stonewall's tactics in this campaign have led to rifts with some gay and lesbian rights activists, and conflict with feminist campaigners and advocates for freedom of expression. Stonewall promotes the twin slogans 'Trans Women Are Women' and 'No Debate', alongside a definition of 'transphobia' which disallows discussion of sex as a real and potentially important characteristic.
Approximately 250 government departments and public bodies, including the ONS, are members of Stonewall's 'Diversity Champions' scheme which promotes gender self-id. The consequence is that self-id policies have been enacted in practice, though not in law. Stonewall has been central to the campaign to erase the status of sex in law, language and data collection. 34 Concerns have been raised that association with Stonewall has led organisations into potentially unlawful and discriminatory positions. 35 31 ONS, 'Correspondence between ONS and the Methodological Assurance Review Panel regarding guidance proposed to accompany the sex question in the 2021 census', FOI request, 17 March 2021; https://www.ons.gov.uk/aboutus/transparency andgovernance/freedomofinformationfoi/corres pondencebetweenonsandthemethodological assurancereviewpanelregardingguidance proposedtoaccompanythesexquestion inthe2021census (accessed 16 June 2021). 32 MBM Policy Analysis, MBM review of the ONS paper 'Methodology for decision making on the 2021 census sex question concept and associated guidance', 8 March 2021; https://murrayblackburn mackenzie.org/2021/03/08/mbm-review-of-theons-paper-methodology-for-decision-making-onthe-2021-census-sex-question-concept-andassociated-guidance/ (accessed 16 June 2021). 33 In fact, the report is not even accurate in listing the organisations that attendees came from. For example, ESRC is listed, whereas in fact no research funding bodies were present. The sloppiness of this report is quite remarkable, particularly in the context of the estimated £900 million cost of the England and Wales census. Jones and Mackenzie describe the behaviour of lobbyists engaging in policy capture as follows: 'Their engagement with policymakers often resembles a process of "co-creating" policy, in contrast to a more traditional model of external lobbying'. 36 This captures the way Stonewall has embedded itself not just within ONS, but within other bodies which ONS in turn consults, including EHRC and the Government Equalities Office (GEO). 37 FOI responses show friendly exchanges between Stonewall and ONS, a quick turnaround of emails, and ease in securing meetings. This is in contrast to the experiences of the signatories of the 2019 letter of concern. 38 For example, the CEO of Stonewall had three personal meetings with Iain Bell in 2020. 39 Emails exchanged prior to the ONS round table show Stonewall suggesting additional invitees, and cautioning: 'I'd strongly recommend that careful thought is applied to how the ONS can practically ensure that, depending on who is present at the roundtable, this is not a hostile environment for trans attendees', to which ONS responded: 'I fully empathise with your words around ensuring that a hostile environment is not created and I can reassure you that ONS will consider very carefully how the meeting will be conducted-I will also arrange for a colleague to take a call with yourself ahead of this event to run through your thoughts or any concerns you may have.' There were in fact no trans attendees present at the event. 40 ONS is a Stonewall Diversity Champion, and a member of the Stonewall Workplace Equality Index. Member organisations make an annual submission and are scored by Stonewall under various criteria, and given advice on how they can improve. The score organisations receive translates into a ranking, with employers competing to be in the Stonewall top 100. Essentially, organisations pay to be lobbied and trained in line with the view that gender identity should always supersede sex. 41 Stonewall has pressured employers to silence or sanction feminist critics. 42 FOI requests regarding this aspect of the relationship between ONS and Stonewall have met with claims by ONS not to have kept copies of their submissions or correspondence: 'The ONS last made a submission to the Stonewall Workplace Equality Index during 2018. We do not hold a record of the submission due to staff changes since then. The ONS does not hold any correspondence that relates to conditions the organisation was expected to fulfil or any policies the ONS was required to change to meet any standards.' 43 These claims to have no record of documentation regarding ONS's relationship with a major lobbyist would seem to represent exceptionally poor record keeping at best.

Application for judicial review
ONS published guidance to accompany the sex question on 12 February 2021. The census date was 21 March 2021. If ONS considered that the short window between its announcement and the census date would preclude legal action, it miscalculated. FPFW brought an application for judicial review of the lawfulness of the ONS guidance, which was heard on 9 March 2021. The claimant, a small, grassroots feminist organisation, raised £101,000 from over 3,000 contributors within just fourteen days. The defendants were the UK Statistics Authority and the Minister for the Cabinet Office, Michael Gove. FPFW applied for 'interim relief', in the form of an order which would remove the guidance until the judicial review was heard. Mr. Justice Swift's verdict granted both permission to apply for judicial review and interim relief to FPFW.
The sex question in the census simply reads 'What is your sex', with response options 'Female' or 'Male'. The contested part of the guidance was as follows: 'If you are considering how to answer, use the sex recorded on one of your legal documents, such as a birth certificate, gender recognition certificate, or passport.' FPFW contested that the inclusion of passports allowed the possibility for respondents to respond in terms of their gender identity rather than their legal sex, while the phrase 'such as' introduced further ambiguity. The introduction of passports as a potential sex marker is problematic given that the sex on one's passport can be changed without a change of legal sex, and indeed, in February 2021, the UK Passport Office stated that it was unable to say how many records have been changed in this way. 44 Legal sex is only recorded on birth certificates and GRCs.
I provided an expert witness statement in the case, outlining the importance of sex as a category in social and medical research, and stressing that sex is distinct from gender identity. Notably, ONS did not provide any expert witness statement to suggest support for its alternative perspective within the quantitative social science community. The legal arguments centred on the definition of sex for the purposes of the census. The Census Act 1920 is the source of legal authority for the census, and gives sex as a permitted particular (that is, information which may be requested). The Census (England and Wales) Order 2020 set the particulars for the 2021 census, including the addition of a voluntary question on gender identity: 'Is the gender you identify with the same as your sex registered at birth?' Justice Swift considered that this indicates a distinction between two separate particulars: sex and gender identity.
Counsel for ONS, Sir James Eadie QC, made three central arguments. First, guided by Iain Bell's witness statement, he asserted that sex is not a simple matter, but an 'umbrella term'. As well as biological sex and legal sex, the term 'sex', it was claimed, may also refer to a person's 'lived sex' or 'gender identity', or to the sex marker on any document issued to them by the state, such as a passport. This entailed the claim, not just that sex is currently 'an umbrella term', but that it was intended as such by the 1920 legislation. The judge dismissed this argument as unpersuasive.
Second, ONS submitted that requiring information on respondents' legal sex risked a breach of the European Court of Human Rights (ECHR) Article 8 on respect for private and family life. It is a legal requirement for UK residents to comply with the census and not to provide false information, and the sex question is a compulsory item. Judge Swift ruled on this point that it was unlikely that there was any such breach, but that if it were, it would be justified, as the question would be posed in pursuit of a legitimate objective. He further noted the careful and confidential way in which census information is used. His ruling on this point was in line with the opinion commissioned by WPUK by Aidan O'Neill QC.
Third, ONS argued that it was not necessary to guide respondents to answer the sex question in terms of their legal sex, as this information could be reverse-engineered by combining responses to the sex question and the gender identity question. Given that the gender identity question was a voluntary open-text response question, the success of such an exercise would inevitably have been partial at best. At any rate, Justice Swift concluded that 'what the ONS suggests seems to me to be like an exercise where a cart tries to pull a horse. I can see little benefit in a state of affairs which requires such a task to be performed.' 45 The law is a blunt instrument, and it seems almost perverse to give a judge authority over a matter of survey methodology. Yet, when public sector bodies fail to act in a transparent and accountable way, the courts are a last recourse. ONS is not the first public body to face successful legal action following capture by the gender self-id lobby. 48 While the status of sex for the purposes of the England and Wales census is now settled and the Northern Ireland Statistics and Research Agency (NISRA) duly followed suit, National Records of Scotland have delayed the Scottish census until 2022, and have not confirmed whether they will revise their guidance on the sex question in light of the judicial review result. Scotland's Chief Statistician, Roger Halliday, has released draft guidance stating that public bodies should never routinely collect data on sex. 49 This extremist position has been challenged by quantitative social scientists. 50 Throughout the UK, and indeed internationally, data collection on sex continues to be undermined by advocates for gender self-id. The claim that gender identity should be collected in preference to sex appears to rely on an implicit assumption that gender identity is more important in determining (all) relevant outcomes than sex. But no empirical evidence has ever been provided to support such a proposition. Indeed, where evidence is available, it suggests that biological sex is the more salient variable. 51 To test the hypothesis that gender identity matters more than sex in any given context, we would need data on both sex and gender identity.
Finally, in a post-fact and post-truth world, and in the face of a global pandemic, the importance of upholding public trust in statistics is greater than ever. Researchers have a well-recognised ethical duty of care towards participants in research, but we also have ethical responsibilities to the readers and users of research evidence, which include doing our best to provide accurate information. If we allow lobbyists to erase a category as fundamental as sex to suit their political narrative and interests, we undermine the ability of social statistics to serve both truth and justice.
Alice Sullivan is Professor of Sociology and Head of Research at the Social Research Institute, University College London (UCL).