Where next for the Census (Amendment) (Scotland) Bill?

people walking on spacious concrete square

This note reflects on the arguments put forward during Stage 1 of the Census (Amendment) (Scotland) Bill and the choices facing the Committee at this stage.

[Numbers in brackets refer to notes at the end.]


The Bill has been introduced to change the law so that the 2021 census [i] in Scotland can include new, voluntary questions to gather information about sexual orientation and, separately, about the transgender population.

In addition, Scottish Ministers are using the Bill to amend how “sex” is defined in the Census Act 1920, to include “gender identity”. This has brought out that Scottish Ministers wish to use a different interpretation of sex for the census than is in current law (mainly the Equality Act 2010) [ii].  It is this which is controversial. It would mean that while the 2021 census would have a question termed “sex” it would not produce data on sex on the usual legal definition for the whole population: for an unpredictable number of people [iii] it would gather a different class of information (gender identity) which would be different from their sex as defined in law, and would not necessarily be limited to two possible responses [iv]. As introduced, the Bill would also set a precedent for the interpretation of sex in primary legislation which conflicts with that used in the 2010 Act and elsewhere [v].

Why this is being discussed

The issue is how the government should respond to the argument that it is wrong to ask some people to disclose the sex recorded on their birth certificate as part of the census.

Some people feel very strongly that the sex recorded on their birth certificate is incorrect and so oppose any requirement being placed on them to report this in the census. Evidence to the Committee reports that some people would find doing so distressing, humiliating or dishonest: some also raise concerns about privacy, although proposed changes to the census process may assist there[vi]. Some people define themselves as having the opposite sex to what is on their birth certificate, and some as being neither male nor female.

It appears to be generally accepted in evidence to the Committee that people who object to being required to declare their birth certificate sex are a different group from those with Disorders of Sexual Development (DSD, sometimes also called “intersex” conditions), although a few people fall into both categories [vii].

Declaring sex as recorded on a birth certificate is most sensitive for people who feel that their original birth certificate is wrong but who do not have a Gender Recognition Certificate (GRC). A GRC allows someone to change the sex on their birth certificate from female to male, or vice versa. Around 5,000 people UK-wide have a GRC. GRC holders are a small proportion of the trans population, which the UK government estimates may be between 200,000 and 500,000 in the UK. There is a lack of reliable data, however. New questions on the trans population in the 2021 census in Scotland, and in England and Wales, are intended to address this.

Obtaining a GRC requires a psychological diagnosis of gender dysphoria, evidence of having lived for at least 2 years in the person’s “acquired gender” (a panel assesses paperwork provided in each case: applicants are not called to interview), a statutory declaration of intention to continue in the acquired gender until death and a fee of £140. A report of medical treatment is also needed, but there is no legal requirement to have had surgical intervention or other specific physical interventions: in this respect the Gender Recognition Act 2004 was seen at the time as world-leading.

Different definitions of sex

The census question on “sex” has been included in the census since it began in 1801 and is compulsory. This reflects that it has always been seen as important data to gather. This appears to continue to be the view of Scottish Ministers, as the Policy Memorandum to the Bill does not discuss ceasing to ask about sex, or making the question voluntary.

Stage 1 evidence has brought out the different types of information the question on sex might be used to collect. These are:

  1. Biological sex: whether people have the type of body that has a female or male reproductive system [viii], which is correctly observed and recorded at birth for all but an extremely small number of people; or
  2. Legal sex (i.e. sex shown on current birth certificate): for most people this is their biological sex, but people who have a Gender Recognition Certificate (GRC) are treated in law as having the opposite sex, in most circumstances [ix]; or
  3. “Self-defined sex”, as female or male: which of these people feel best describes them at the point they complete the census, regardless of what is on their birth certificate at the time. This merges sex with the separate idea of “gender identity”, so that, as described above, the question would gather a mix of data on sex and on gender identity; or
  4. “Self-defined sex”, not limited to male or female: As above, but this allows people to record their legal sex, the opposite sex, or to define themselves as a third category which is something other than female or male (for example, some people who describe themselves as non-binary are seeking this). This also merges sex with the idea of gender identity, but more completely, and was Scottish Ministers’ preferred approach when the Bill was introduced, but it is less clear at this stage how far that remains the case [x].

The difference between options 1 and 2 is currently so small in absolute numbers (probably around 500 people in Scotland) that the choice would not be expected to make any difference to the nature of the data collected in the census, although this depends somewhat on any changes to the GRC rules before 2021[xi]. Also, it is arguable that GRC holders are legally protected from having to disclose their biological sex [xii]. It is therefore assumed below that option 1 can be ignored here and that for the purpose of the census option 2, “legal sex”, provides a good enough proxy for biological sex, while avoiding any clash with legal privacy rights.

The relevance of how sex is collected in the census

The census is used to build the most accurate possible description of the characteristics of the Scottish population, including how these relate to each other (for example, the relationship between sex, age, education and employment). It allows the results to be broken down into small groups, by geography, age or other features. Census data is a centrally important input to national and local strategic planning of public services, to policy development and to research into how society works [xiii]. Census data also contributes to monitoring at a population level of the effectiveness of the Equality Act, under which “sex” [xiv] is one of nine “protected characteristics”.

Some evidence to the Committee argued that it has been commonly understood for many years that having female or male biology directly explains some differences within the population which are relevant to the use made of census data (e.g. in estimating birth and mortality rates or population-level susceptibility to certain medical conditions); and also that women and men’s social experiences (for example, in employment) may be different because of differences in the way people tend to be treated according to their biological sex.

However, some submissions to the Committee propose that the only information the census needs to gather is on the incidence “self-defined sex” or “lived identity”.  This assumes that differences between people which will be relevant to the users of the census data are better explained by whether people perceive themselves to be male or female (or neither), and how that is then reflected in how they live their day to day lives, than by their biological sex, or their legal sex.

In other submissions arguing for the use of “self-defined sex”, it was not always clear whether data on biological/legal sex was regarded as unnecessary in principle or whether the numbers self-defining as other than their biological/legal sex are expected to be too small to matter in practice [xv].

Some evidence to the Committee argued that if biological/legal sex is needed, it could be worked out by cross-referencing a self-defined sex question with the answer to a separate question on whether or not someone is trans. That would rely on certain assumptions being borne out and for technical reasons would work least well with the government’s preferred option on introduction [xvi].

Other evidence argued that data should be collected both on biological/legal sex and, voluntarily,  on trans people’s self-defined identities, to maximise what we can learn from the census with minimum reliance on inference.  Having looked at the evidence gathered at Stage 1, this might be considered the most pragmatic solution. Gathering both data sets would mean reliable data was available on sex as a protected characteristic, and interactions between sex, gender reassignment and other protected characteristics could be reliably examined in different contexts.  This approach also makes the fewest assumptions about whether biological/legal sex or lived identity, or both, will be the most useful population-level information to have in the many different contexts in which census data are used. This is important if, as it appears, substantial work remains to be done with a broad range of census users to establish their perspective on this choice.  It may now be getting late to do that, and in any case until they have been able to compare the results of the two data sets in their own area for the first time, some census users may be unwilling to make that judgement.

In summary, the debate here has been whether:

  • Legal sex, as a close proxy at present for biological sex, continues to be an important characteristic to know about in the population, overall and in combination with other characteristics.
  • If so, whether it needs to be gathered directly; or if the number of non-GRC holders self-defining as other than their legal sex could be relied upon to be so small that the fuzziness introduced into the data would not matter; or whether legal sex could be inferred reliably enough from cross-referencing “self-defined sex” and questions on being trans.

It is not absolutely clear what the government’s position is on these specific points, other than it does not believe legal sex needs to be gathered directly, and what evidence and analysis of how census data is used underpins whichever position it takes. This is a point the Committee could ask the government to clarify.

Consistency with other data

Some evidence to the Committee has stated that in various contexts, there has already been a move to record “self-defined sex” or “gender identity” rather than sex as recognised in law (e.g. in the last Scottish census for those aware of new on-line guidance, in the Scottish Household Survey as of 2018, in NHS patient records, as explained here and by various other public bodies). It is not clear whether any of these recent decisions have been previously reported to the parliament.  Ceasing to collect sex as defined in law raises questions about the capacity of organisations to comply with the Equality Act 2010 and so ought always to be subject to an Equality Impact Assessment on the characteristic of sex [xvii].  The issue brings out the challenge of balancing individual privacy/respect for self-defined identity and fulfilling Equality Act responsibilities in relation to the protected characteristic of sex.

Because of the importance of the census, whatever is done here is likely to have a large influence on whether the implications of the recent trend among public bodies to move away from collecting data on sex as defined in law will be open to debate, or will be consolidated and extended.

For the census in England and Wales in 2021, this statement was recently issued. The government here has told the Committee it is not yet clear whether or not the ONS’ intention is to ask for information on legal sex.

‘Maintaining the current question is important to preserve continuity of data in respect of the protected characteristic of sex. The question on sex (male or female) is established in the census, and it is essential to the evaluation of inequality related to that protected characteristic. Consideration has been given to amending the question to reflect a wider range of options, given that there is greater recognition than previously of individuals who reject the traditional “binary” view of sex. Nevertheless, the protected characteristic of sex as defined in the Equality Act 2010, and as relevant for the PSED, is whether a person is a man or a woman. This binary concept of sex is, in turn, fundamental to the Equality Act 2010 definition of sexual orientation and of gender re-assignment, and to the law on marriage and civil partnership and many other matters.’ (Office for National Statistics, 2018)

Alternative approaches

If MSPs took the view it was wrong to require certain people to record the sex on their current birth certificate in the census, an alternative to moving to “self-defined sex” would be to ask for legal sex, but to include some form of “prefer not to respond” answer.  The government evidence does not discuss this option. It would raise questions about reduced data quality, although these ought to be no worse, and potentially less acute, than those raised by using “self-defined sex” [xviii].

If legal sex is collected with or without a “prefer not” option, the accompanying advice might  still acknowledge that this question will not be straightforward for some respondents. It could stress that the question on sex is only collecting data on the fact of current legal status, as this is valuable information to have at population level for everyone, including the trans population, that nothing will be assumed from it about how people identify, and that trans people will have the opportunity to record that in a separate question. This would require the questions for trans people to provide that possibility.

Issues for the Committee

(i) Decisions for the Bill text

The proposed amendment to include the concept of “gender identity” within “sex” opens the way to the question on sex allowing options other than female and male, and to setting a more general legal precedent that “sex” can be understood other than as established in law at present, to have more than two categories. The government appears to be willing not to make this amendment and has said it will take the Committee’s advice. So the Committee will need to decide whether it agrees the Census Act 1920 should be amended in this way (accept the amendment to Paragraph 1 of the Schedule to the 1920 Act) or not (reject the amendment).

The Committee will also need to agree the wording used to pave the way for one or more questions gathering information on the trans population. The Bill currently suggests “gender identity” but again the government appears not to be sure this is the best option and witnesses from all perspectives have raised questions about the phrasing of this part.  There is a good case for choosing whatever wording leaves the maximum room for further question development in this new area for the census: for example, avoiding a revised wording which allowed a question about whether a person was trans but not any question about their lived identity. There is an argument for aligning at least one part of the question with the Equality  Act protected characteristic of “gender reassignment”, to assist monitoring the effectiveness of that legislation, as suggested in some evidence to the Committee.

(ii) Possible points to make  in the Stage 1 report for policy beyond the Bill

It is not clear whether the government believes a “third sex” option could still be included in the census, even if the reference to sex in the Census Act 1920 is left as it is. Therefore, if the Committee does not agree the census should go beyond the two sexes which exist in law, it may need to make that clear, over and above whatever position it takes on the wording of the Bill.

Separately from considering the number of sexes the census should recognise, the Committee will have to decide what to say in the Stage 1 report about the choice between legal sex and “self-defined sex”.  This is not an issue for the drafting of the Bill. Based on the evidence it has received, it could offer a view on which it thinks should be collected, or whether it believes both should be, and/or whether Scottish Ministers need to do any more to allow informed discussion and decision-making about this choice, and if so, what sort of process and evidence it would expect to see before the government makes a final decision.

If the Committee sees a case for collecting data on legal sex, but also feels that people should not be compelled to record their legal sex in the census, it could also ask Scottish Ministers to consider the implications of including a “prefer not to respond” option under “sex”.

On all these points, the Committee is being asked to judge how the reliability and usefulness of the data collected in the census is best maintained for users while recognising specific concerns raised by some of those who will be asked to fill it in.


[i] The census takes place every ten years. It is the only time that data is gathered systematically from the whole population across a range of characteristics.  Other data sets are sampled (e.g. the annual Scottish Household Survey) or are gathered for administrative use for a specific function (e.g. NHS data). Although it does not achieve a 100% response rate, compared to other sources of information the census still provides more reliable information even when broken down into relatively small units, such as narrow age bands, or geography.

[ii] The established definition in law of sex is as being either female or male only, with the Equality Act 2010 being the most relevant primary legislation: “In UK law, sex is understood as binary” (Equalities and Human Rights Commission). Evidence to the committee has shown there to be some dispute about exactly how sex is defined in law. However, there appears to be a consensus that it is limited to female or male and based on some observable characteristic (either sex at birth or as shown on current birth certificate), and is therefore distinct from gender identity, which is self-declared and not limited to two categories.

[iii] The practical effect might be minimal, but this is impossible to test on available information. If the effect of the change were concentrated on particular parts of the population (such as particular age groups), small numbers nationally could still have a substantial effect on the analysis of the data specifically involving those sub-groups.

[iv] The introduction of the Bill has brought out that an initial move in this direction was made in Scotland in the 2011 Census, as set out in more detail here. The 2021 census plans appear to be the first time that this decision has been opened to scrutiny. The 2011 change was narrower in effect than the one proposed for 2021, as it retained a two-category definition of sex.

[v] Although in almost all legislation sex follows the model in the note above, one alternative legal precedent exists in the Gender Representation on Public Boards (Scotland) Act 2018. An amendment made during Stage 2 to Section 2 inserted that ““woman” includes a person who has the protected characteristic of gender reassignment (within the meaning of section 7 of the Equality Act 2010) if, and only if, the person is living as a woman and is proposing to undergo, is undergoing or has undergone a process (or part of a process) for the purpose of becoming female””.  This appears to have the effect of bringing trans people who may still hold a male birth certificate within the scope of the legislation provided they are transitioning to live as a woman, and excluding from scope all other trans people, including any holding a female birth certificate issued at birth, whether they identify as men or something else. The legislation did not require a mirror image definition for “man”.  The existence of this precedent has been cited in evidence to the committee by organisations which led on promoting for the amendment in 2018 as a reason to change the way sex is defined for the census. However, the very narrow scope of the 2018 Act, its provision of definitions for “woman” but not for “man”, and its implied lack of clarity about how to define in law some trans people (those transitioning to live as neither a woman nor a man) make the 2018 Act a much less obvious relevant legal precedent for the census than the Equality Act.

[vi] Unlike, say, a passport, the census return is not a public document, but it is normally completed by household, so that information returned may be seen by at least one other person in the household. From 2021, the SG is planning to allow people to apply for a separate individual return, to increase the protection for privacy. If any data processors see name and sex together, disclosing any such information carries substantial penalties under census legislation. The census data is mainly used in an anonymised form.

[vii] This submission to the Committee from a DSD charity states that most DSD cases are disorders of either a female or male reproductive system, meaning that most people with DSD conditions belong, and recognise themselves as belonging, clearly to one or other sex.  It estimates around 12 babies may be born in Scotland each year  who will need additional tests and specialist input to establish sex, but that in most of these cases there will still be a clear basis for a female or male decision . It estimates that there are around 7 or 8 cases born in the UK each year who are truly hard to categorise as either female or male, but not that these are a distinctive third sex requiring (or wishing for) a separate census category.   Of these cases, a small number will at some later point decide that they do not accept the sex initially recorded for them: DSDFamilies notes there is no systematic data on the number of people with DSD who decide that they wish to change the sex assigned them at birth, but estimates it may be 1 born in the UK every two years, or 1/260 people born with a DSD condition.

[viii] Whether a person is reproductively female or male is established in humans before birth (though see footnote above). There is some debate about how the impact of any later hormonal treatment and surgery should be described. Depending what is done, later changes may mean that to a greater or lesser extent some birth sex-based physical issues cease to be relevant (e.g. if organs are removed) and some others more associated with the opposite birth sex become more relevant (e.g. due to taking cross-sex hormones), but such changes cannot move someone from one reproductive type to the other.

[ix] For example, a Gender Recognition Certificate does not allow a person born female to obtain rights to inheritance down the male line (property or titles). There is also some disagreement about whether people who have used a GRC to change their legal sex have full access to single sex spaces/occupations/activities under the Equality Act 2010.

[x] The National Records of Scotland has said that where people respond that they are something other than female or male it will then assign these cases to one or other sex for the purpose of any detailed analysis of census data, only reporting the number of “other” cases at the aggregate national level.

[xi] The characteristics of the GRC-holding population are recorded by age and direction of female/male change. This means that reliable information on biological sex could still be derived from legal sex at the aggregate national level. An increase in GRC holders could happen under Scottish Ministers’ plans to simplify the process for obtaining a GRC by removing the existing criteria of medical diagnosis and a period of lived experience, but on what scale is unknown. If the number of GRC holders were to increase in any substantial way before the 2021 census they would remain a “known unknown”, although as the data were broken down further its usability as a proxy for biological sex would be more affected.  The Equality Network has suggested that experience in other countries would suggest that the additional take up of GRCs would not be large if the rules were simplified.

[xii] The Gender Recognition Act 2004 contains provisions limiting when a person’s possession of a GRC can be disclosed. These were intended to give effect to privacy rights under the European Convention on Human Rights.

[xiii] Evidence to the committee noted that in Scotland, census data is used as a main mechanism for linking together anonymised data from other sources, to produce more detailed information on how particular characteristics in the population relate to each other, and that Scotland is at the forefront of such data linkage, in which sex is one of the main variables used to connect records.

[xiv] There is some dispute about whether it is biological or legal sex which the Act protects.

[xv] It is difficult to know what the take-up of self-definition would be, as this was not measured in the 2011 census and does not appear to be available from any other source. The UK government has estimated that up to 1% of the population may be trans. However, it is not clear how far the percentage could be substantially higher in certain sub-groups (for example, by age), which could mean self-definition could have stronger effects on the data for those groups. If there is an association with being younger, it would affect the data for the group which already has the highest non-response rate in the census.

[xvi] The response “other sex” combined with a Yes/No response on being trans would give no information on biological sex.  Even with option 3, inferring legal sex assumes that all trans people will define as the opposite sex: some evidence to the committee noted that an unknown proportion do not change their description of their sex, even if they live in a new “acquired gender”.

[xvii] The Act requires public bodies to monitor and report the impact of their activities on the legally protected characteristic of “sex”. Organisations are also enabled to provide single sex services, occupations etc, which requires them to know the sex as defined in law of service users or employees, as appropriate.

[xviii] The number who would opt for “prefer not”, for whom no data on legal sex would be collected directly, is hard to predict. However, the impact on data quality is potentially less than with self-defined sex. On a “prefer not” option it is clear which cases have chosen not to provide legal sex, and conversely that any data which is provided on sex reliably gives legal sex. If combined with a separate question which allows trans people to record their gender on their own terms, it ought also to be possible to make a reasonable inference of legal sex in many cases.