Letter to the CTEEA Committee: key issues on the census sex question (11 September 2019)

Letter to CTEEA Committee, setting out key issues and concerns about the sex question in Scotland’s census.

11 September 2019

Dear Committee Members,

We note that as part of its preparation for the 2021 Scotland census, National Records of Scotland (NRS) are planning to commission testing which looks at the effect of offering different forms of accompanying guidance for the sex question.

The guidance is important because this has the effect of determining the meaning of the sex question. Currently, NRS propose to publish guidance which will indicate that sex has no definition beyond the respondent’s self-assessment (self-identified sex). We understand that the Office for National Statistics (ONS) are proposing similar guidance for the England and Wales census. However, recognising that concerns have been raised about gender self-identification as a basis for answering the sex question, the guidance will be subject to further testing over the coming months.

Testing will compare the results of the sex question having accompanying guidance that will indicate that sex has no definition beyond the respondent’s self-assessment; with guidance that directs people to answer in terms of their legal sex, as recorded on a person’s current birth certificate or their Gender Recognition Certificate (which applies to around 400 to 500 adults aged 18 years or over in Scotland). Both sets of guidance will be tested in conjunction with the trans identity question (which is explicitly designed to collect data on gender identity), and the sexual orientation question.

This will look at any variation in response rates to the question between the two approaches, how respondents interpret the question, and if and how responses in each case interact with the voluntary transgender and sexual orientation questions, and data quality.

Online format
We note that the 2021 census will be web-based, and a link to the guidance made available on the online questionnaire. This format means that the accompanying guidance will be both prominent and accessible.

This level of accessibility is much higher than the 2011 census. Online guidance on the sex question was provided for the first time in 2011, which stated that transgender or transsexual people could select the option for how they identify, irrespective of the details on their birth certificate. NRS advise that only 20% of households responded online in 2011, and even for those respondents, the guidance was not linked to the online questionnaire.

This meant that respondents, whether on paper or online, needed to actively seek out the online guidance. We have seen no record or assessment of how far the 2011 guidance was used by respondents, or how this influenced responses. The difference in approach between the two censuses however makes it a reasonable assumption that any guidance used in 2021 will be seen much more widely than guidance published in 2011.

Key issues
We are concerned that the NRS approach to testing does not address any of the key concerns in respect of a self-identified approach to the sex question. These are:

  1. the fundamental redefinition of male and female within the context of Scotland’s oldest and most important population survey;
  2. the precedent this sets for other surveys and more widely; and
  3. the departure from the legal definition of sex in the Equality Act 2010, which carries implications for equalities monitoring.

From a data users’ perspective, a question that solicits responses based on gender self-identification is not based on clear criteria and instead conflates two separate demographic characteristics, namely sex and gender identity. In other words, it is not a sex question.

This carries significant implications for data quality. It means that for an unknown (and unknowable) number of people, the census will collect a different type of data to sex. With no reliable estimate for the moment as to the size of the trans population, it cannot be safely assumed that this will have no impact on the value of the census as a source of reliable population-wide information about sex, and how it relates to other policy areas, such as health, education and employment. From the limited evidence we have, the risk to data quality is likely to be higher among some sub-groups, particularly young people.

Giving evidence at Stage One of the Census Bill, Professor of Quantitative Criminology Susan McVie stated:

I think that the General Register Office for Scotland got it wrong when it redesigned the census in 2011 and conflated sex and gender identity into one question. We are now trying to disentangle those things. Arguably, the measure of sex in the 2011 census data is not accurate. (McVie, 13 December 2018. Col. 4)

A sex question soliciting answers based on self-identification is also not consistent with the Culture, Tourism, Europe and External Affairs Committee (CTEEA) recommendation in the Stage One report on the Bill, which states that the approach must ‘[take] into account the census’s primary purpose of robust data-gathering and the Scottish Government’s duty to act in accordance with the Equality Act 2010, in which sex is a protected characteristic. (CTEEA, 2019: para. 8).

It is unclear why NRS are proceeding to test guidance for the sex question soliciting responses based on self-identification. Cabinet Secretary for Culture, Tourism and External Affairs Fiona Hyslop MSP has stated that this approach will maintain continuity with the 2011 census. However, the guidance published in 2011 was not only much less easily available than any guidance will be in 2021, it was also an outlier. No guidance related to sex was published in the previous censuses between 1801 and 2001.

The introduction of a separate transgender status question in the 2021 census to record gender identity also means that the 2011 precedent is no longer relevant. Nonetheless, it appears that the Scottish Government views the previous fudge as a permanent concession that respondents should never declare what is on their birth certificate. It has not however, made clear why this concession is justified, nor how it reached its conclusion.

The primary purpose of the census is to provide information to develop policies, plan and run public services, and allocate funding. The data collected in the census should reflect user needs, be of high quality and serve the public good. We have not however, seen any robust arguments to support the position that a sex question soliciting answers based on self-identification would be consistent with these aims and principles. NRS have acknowledged that they have not been asked to follow this approach by any data users. 

We do not believe there is any value in testing the effect of offering guidance based on self-identified sex, because soliciting responses on this basis is not consistent with the primary purpose of the census. It is therefore unclear how further testing could contribute to the decision-making process.

We think that the sex question should, as it states, collect data on sex, and not data on gender identity. Any further testing should therefore have as its clear and sole purpose ascertaining the best way to do this. NRS may wish to look at how people are likely to understand and respond to the sex question, and the value of providing additional guidance based on legal sex. For example, testing might consider the extent to which respondents are likely to use such additional guidance, and whether this is likely to affect how a person responds.  Data on gender identity, where this differs from a person’s legal sex should be clearly collected as a category in its own right, using the separate voluntary transgender question.

The census is a once-in-a-decade opportunity to survey the population. It is a complex and expensive programme (£64 million in 2011) and getting the data right matters. In this respect, we believe that the Scottish Government should continue collecting reliable data on sex until there is compelling evidence that it is no longer needed.

Kind regards,