Culture, Tourism, Europe and External Affairs Committee

Meeting date: Thursday, December 13, 2018

Official Report 703KB pdf

---

Census (Amendment) (Scotland) Bill: Stage 1

Good morning, and welcome to the 33rd meeting in 2018 of the Culture, Tourism, Europe and External Affairs Committee. I remind members and the public to turn off their mobile phones. Any member who is using an electronic device to access committee papers should ensure that their device is turned to silent.

Agenda item 1 is consideration of the Census (Amendment) (Scotland) Bill at stage 1. This is the second meeting at which we will consider the bill. The purpose of today’s meeting is to consider the gathering of voluntary data on gender identity and sexual orientation.

We have with us a panel of data users and analysts, whom I thank for coming to give evidence today. Lucy Hunter Blackburn is from Murray Blackburn Mackenzie; Professor Jackie Cassell is the head of the department of primary care and public health, and the director of research and knowledge exchange, at the Brighton and Sussex medical school; Gerry McCartney is the head of the public health observatory division at NHS Health Scotland; and Professor Susan McVie is a professor of quantitative criminology at the University of Edinburgh.

We were also due to take evidence from Ipsos MORI. However, because of a lack of videoconferencing facilities in the Ipsos MORI office, we are unable to do so.

The main purpose of the bill is to add voluntary questions on gender identity and sexual orientation to the 2021 census. In the course of scrutinising this short bill, the committee has received submissions in which concern has been expressed about the conflation of the categories of sex and what is called “gender identity” in the bill. In a number of those submissions, it has been suggested that the approach could be problematic for the gathering of data and could set a precedent.

In “Scotland’s Census 2021—Sex and Gender Identity Topic Report”, the census team said:

“Sex is a key demographic variable”

and that sex data

“is a vital input to ... demographic statistics which are used by central and local government to inform resource allocation, target investment, and carry out service planning and delivery.”

The Women’s Budget Group has said that it relies on sex-disaggregated data in analysing the economic and social impacts of public policy on women.

Lucy Hunter Blackburn, you express the same concerns in your submission. Will you talk about them, particularly in the context of data gathering?

The starting point is the purpose of the census, which is to gather data that is as useful and reliable as possible. By “reliable”, I mean that we get the same kind of answer from the same kind of person, so we know what the answers mean and people who plan public services can use them. Above all, the census exists for the specific purpose of data gathering.

The main point that I want to make before I say anything else is that there is clearly no disagreement at all about the value of gathering more data on the population who identify as trans. The gathering of such data is immensely valuable, and the census is a good place to do that. The insertion of voluntary questions on that issue does not seem to be a matter of contention here; really, the debate is all about what happens with the question on sex.

In that regard, let me make three points. First, it is important that the census continues to capture data on sex, as a protected characteristic, in terms that are consistent with the Equality Act 2010. The 2010 act clearly has a two-category definition of sex: male and female—there are no more categories. To introduce a third sex category, as the bill proposes that the census be amended to do, would take the approach out of consistency and compliance with the Equality Act 2010, given how it is framed. That would definitely create issues for quality monitoring and data use.

My second point is that there is no evidence that service users have demanded a move away from binary sex categories in the census. The National Records of Scotland is clear that the proposal is not driven by census users wanting a third category.

A third, and different, point, which is also important, is that you need to avoid using the census legislation to set a precedent on the statute book that conflates sex with gender identity. Those are distinct things, conceptually, and we do not feel that the census legislation is the right place to introduce new ideas about how to conceptualise sex in law. That debate could be had, but not in this context.

Our view is that the definition of sex in the Census Act 1920 should remain as it is now, so that it is in line with the definition in the Equality Act 2010 and in other legislative contexts where sex is used, such as marriage, birth and death certificates. We use a two-category version of sex right across the legal context. That suggests that we should not amend paragraph 1 of the schedule to the Census Act 1920, which talks about sex, by introducing the proposed additional words, which would mean that we would need to introduce gender identity in the bill in the same way as it introduces sexual orientation—as a voluntary topic.

Thank you. I should say, for the record, that there will be either a gender identity or a trans question, the wording of which we will consider later, in addition to the sex question. Do other members of the panel wish to comment on that point?

I will comment on the use of such data. There are many uses of data on sex, some of which are more important than others, and we should look at the wider context of that data. It is important that the census is a point of reference. With the growing use of administrative data sets—a couple of censuses away, the census will look very different and we will draw on national health service data and all sorts of data—it is important for the credibility of the census as a data resource that gives information at a low level on small populations that there is consistency between the census data on sex and other data sets.

The various routine collectors of sex data have public equality duties, and it is important that people do not start looking elsewhere because the census data is seen to be problematic. There is a real issue about precedent and the credibility of the census and, for the purposes that those many data sets are drawn on to provide sex data, it is key that we have a good representation of the definition as it currently stands in law. That can, of course, be discussed and moved on, but it is important that the census is consistent with other data sets and with the definition as it stands now.

What is the importance of that for health purposes?

There are many areas of health in which biological data and biological risk are very important. Those interact with gender identity, which is very important in some areas. We know that, for many conditions and treatments, sex is a big differentiator of outcomes. It is important that the data is robust in order that we can meet our duties, which are many and are often not well met, in the fair provision of effective treatment to men and women, which may not always look the same.

It is important that the committee differentiates between two different things: the provision of services for individuals and population-level data. Clearly, we do not use census data to institute or identify services for individuals. Notwithstanding the points that have been made, we would not use the census to identify people for screening or other such purposes; we would use the existing health record data sets such as the community health index data sets and the clinical records associated with those to identify individuals who needed particular services. However, the census is a key data source for resource allocation and planning at a population level. It is important to recognise the difference between those two data sets.

In Scotland, we have some extremely rural areas with small populations, where small differentiations in the census data—or any data—would be significant for resource allocations.

Yes. The census is used for a variety of purposes, including the denominator data that calculates the index of multiple deprivation and other such indices right down to extremely small geographical areas such as data zones. All the data that we collect in the census is used, through other means, to make those more precise resource allocation decisions for local areas.

Professor McVie, criminology is your area. Will there be an impact on that?

I am not here to speak on criminology per se; I am really here as the co-director of the administrative data research centre in Scotland.

Scotland is at the forefront of data linkage, and the census sits at the core of data linkage. We have what we call a census spine. The census forms the core of how we link all other administrative data sets together, so it is vital that the census is accurate in its measurement of the characteristics of the population in order that it can form that strong spine to which we attach other data sets.

It is a fundamental property of research that, in designing a questionnaire, you need to be extremely clear about what you are measuring. Possibly controversially, 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.

In your papers and discussions, you have probably come across the fact that the issue of sex and gender is not simple but quite complex, with lots of dimensions to it. Trying to boil it down to two questions in the census is somewhat problematic, which is probably why we are all around the table today, trying to disentangle the different aspects of it.

From a research point of view, we know that certain conditions—medical conditions, for example—are sex related. Regardless of a person’s gender identity, there are certain medical conditions that they will be more likely to face depending on whether they were born a man or a woman. We also know—this is probably more to do with my area—that certain social processes are differentiated for men and women. There are sex-related biases, discriminations and forms of inequality that do not necessarily go away if a person changes their gender identity.

It is important to distinguish between sex, on the one hand, and gender identity, on the other, in order for us to understand, for example, whether trans women have worse outcomes than cis women and whether trans men have worse outcomes than cis men. If we are to properly understand the relationship between sex and gender identity and how that impacts on factors such as health, the likelihood of getting a job and attainment in education, we need to disentangle those things so that we can have a much clearer picture.

You say that the GROS got it wrong when it conflated sex and gender identity. It has been put to us that there were notes that explained that; however, when we asked for a copy of the census form, we found that the guidance notes were not on the census form itself but were online somewhere. Do you think that people understood that sex and gender identity had been conflated?

Some people would have gone to the trouble of reading the guidance notes and would, therefore, have understood the situation, but other people would not have done that. The problem is that some people would have interpreted the question as asking about their biological sex, whereas others would have interpreted it as asking about their gender identity. That means that we do not know what that question was measuring in relation to any particular individual.

In addition to that more general conflation, there is a proposal to offer a third option on the sex question in 2021, so the answer could be male, female or other. The wording is not yet decided and that is not set in stone, but it is suggested that that might happen. How would you feel about that?

09:15  

I have spoken to members of the lesbian, gay, bisexual, transgender, queer and intersex community, and I have to say, for a start, that the word “other” is highly offensive. Many people from that community do not consider themselves to be “other”. Fundamentally, we are still conflating two things if we use a category of “other”. Sex is about either biological or legal sex—whichever you decide to use—whereas gender identity has non-binary options. Sex does not have non-binary options. Even someone who is intersex, which is essentially a medical condition, is generally an intersex male or an intersex female on the basis of their physical and genetic composition, so you would still be conflating two things if you added an “other” category.

The main thing to consider is what the different dimensions of gender identity are. There needs to be a series of questions that allow people to adequately express how they feel about themselves, keeping sex separate from that and being very clear about what we mean by sex, with guidance in the documentation. There should then be a publicity campaign around the census that explains why the questions are phrased as they are. It is not the purpose of the census to make people choose something that they do not want to choose to represent themselves; its purpose is to measure the characteristics of the population so that the data can be used to properly understand how such things as healthcare, social experiences and education can be delivered. In the era of measuring inequality, we know that many people from the LGBTQI community feel that they are discriminated against, and we will not properly be able to understand how that discrimination manifests itself if we do not understand what their sex is.

I want to ask about the voluntary aspect of the sexual orientation and gender identity questions, as they are phrased at the moment. The Equality Network and the Scottish trans alliance said in evidence that it should not be assumed that adding those questions to the census will give an accurate count of the trans and LGBTI community, and that the questions might be seen as being too sensitive, so some people would be reluctant to answer them. If the questions are voluntary and the data will not be reliable, is it worth asking the questions at all, or will helpful data and information come from the questions being answered on a voluntary basis?

We have some evidence from “Scotland’s Census 2021—Sex and Gender Identity Topic Report” that take-up of the voluntary question was quite high. From memory, I think that something like 94 per cent of people put something in when it was road tested, so the initial evidence is that people will respond. However, others might wish to say more about that, because that report did not tell us what the response rate was among people who were trans identified, and that is clearly where it matters most.

The evidence from three waves of the national sexual attitudes and lifestyle survey is that such questions can get good response rates, even outside the census. It is likely that questions about gender, sexual orientation and trans identity will change over time, because they shift with social change, but there is strong evidence that it is possible to get reasonable quality data on many such aspects from surveys that have been done in many settings with general population samples, which supports the questions’ acceptability.

They are complex questions, and some people will have better understanding of them than others. There is also evidence that there will be reasonably high non-response rates, but there is no doubt, from many studies of many kinds and population samples, that it is possible to get useful data that can inform policy.

The religion question in 2011 was voluntary, but we have made huge use of the data from it and have been able to explore the prevalence of religions across Scotland and to use that information to think about discrimination and other important facets of society.

There are also examples of where we have been able to link data—in a complicated way, to avoid individual identification—to make better use of the ethnicity data from the census. That has allowed us to explore differences in life expectancy and hospital admissions among the different ethnicities. It has really moved forward the evidence base and what we know. For example, life expectancy for white Scots is much lower than life expectancy for many of the ethnic minorities in Scotland, but hospital admissions for some conditions are higher for some ethnic minorities. We would not have known those things without the questions in the census.

I know that the ethnicity question was not voluntary, but we can make similar use of the data even if questions are not completed entirely, because we will still get a feel for the differences between groups.

I am not sure whether you have had a chance to look at the bill, but it is very short. We have already had responses from the panel about the use of the term “gender identity”, which is, as Lucy Hunter Blackburn said, in section 1. It has been suggested that, when it comes to the voluntary question, the term “gender identity” could be changed to “trans” so that the question would be more specific. Do you support that change in wording? Would it work better than the current proposal?

Professor McVie talked about the previous census in 2011 and questioned the robustness of the data from the sex question, given that the guidance was online and was not that obvious to people. Have there been any unintended consequences of the decision that was taken in 2011, or is it too early to make comparisons? Have problems or issues arisen as a result of our not being clear about that question?

I am sorry that this is so long. I have a final question. In the context of the society that we live in, is it possible to ask a question on sex in a way that means that we are clear about how people will answer it?

I will try to remember all the questions. Can we make a question about sex clear? Yes, we can. We can ask about biological sex or the sex on a person’s birth certificate, which, in effect, is their legal sex. That means the sex of someone when they were born or, if they have a gender recognition certificate, the sex to which they have transitioned. That would be clear.

I am trying to follow back your train of thought.

My question was particularly about whether there have been unintended consequences from the question in 2011.

It is impossible to tell, because we cannot disentangle the data if we have conflated two things. We would know that only by getting the questions right in the 2021 census. We can link censuses together, so we would be able to link back to the prior census and identify the proportion of people who interpreted the question as being about their biological or legal sex, and the proportion who identified as a trans man or woman and so put their preferred gender identity. The honest answer is that we do not know the numbers, which is why the census is so important. The census is the only source of data that we have that is an entire-population measure. It is important to measure the population accurately, but it is more important to ask questions that are clearly differentiated so that people understand what they are being asked.

You also asked about the proposed questions.

I asked about the term “gender identity”. Should it be in the bill? Will that make it clear what the questions will be about?

I am not aware that a definition of “gender identity” has been agreed. In the Equality Act 2010, the protected characteristic is “gender reassignment”, and even that is a bit vague, to be perfectly honest. “Trans” and “gender identity” are wider concepts than “gender reassignment”. If you want the bill to meet the Equality Act 2010, the question should be focused on gender reassignment, but if you want to respect the wishes of many LGBTQI communities who want their self-identified gender to be recognised in the census, you will need to ask a wider set of questions. That is why it is difficult.

This is such a complex area and it is very difficult to squash it down into one question about sex and another about trans, gender identity or something else. Unfortunately, there is not yet consensus on what is meant by “gender identity”. There is possibly more consensus on what people might identify as trans, but that is broader than gender reassignment. It is tricky terrain in definitional terms.

Good morning. I want to go back to the 2011 census guidance on self-identification. It has been argued by some that because the cohort of people who followed the guidance and answered the mandatory question on sex accordingly would be quite small, statistically it would not have made much difference to the accuracy of the data. It would be interesting to hear your comments on that argument.

If we look at the 2011 census versus the 2021 one, that is a big issue. We cannot tell what happened in 2011, although, as Susan McVie has said, if we were to get very good clear questions in the next census we could go back and quantify that better. The 2021 census is very much an unknown. I would be very reluctant to assert that numbers would be smaller in 2021 through people taking advantage of the ability to identify as something other than their birth certificate sex. We just do not know.

However, we do know that it is a growing phenomenon: it is emerging that the numbers of people who are presenting in the most formal sense—in gender reassignment and gender identity clinics—is going up very fast. However, that is only one part of the trans population; not all of them will engage with such services. We would be introducing a huge unknown into the data if we were simply to add an extra category, and we do not know how many people would use it. We could not do much if we were to do it simply on a gender identity conflation basis. We could not do all the things that Susan McVie so clearly described—trying to disentangle all the characteristics of a person so that we can really know about them. That is the problem with trying to say that the numbers will be smaller in 2021. I do not think that there is any basis for saying so.

I work largely on data about education and higher education. We need to think not just about the aggregate numbers in the population, but about where they might be concentrated. The data that I use is mainly about people in their teens and 20s who are in full-time education. If the phenomenon of taking advantage of gender identity flexibility in the question were to be particularly concentrated in that sub-group, the data effects would be much more concentrated in that sub-group, so the data would be unlikely to affect people who are aged 50 plus. I would guess that there might be a minimal effect, but as you know, down the population that could change. For me, as a data user, it is important to know that the effects could be very unequal in sub-groups, and in ways that really matter.

That is interesting. Do other panellists have any comments on that point?

I agree with that. We might think that we will have small numbers, but until we have a measure in the census we will not know how small or large the number is. Lucy Hunter Blackburn is absolutely right: for a population at macro level it will probably not make a lot of difference. However, the purpose of the census is not just to carry out macro-level population analysis; it is also for micro-level analysis. That is why we collect such detailed information about individuals and households. We find, for example, that vulnerable and marginalised populations are often very small, but if we do not have accurate data on such groups, we cannot tell how badly discriminated against they are.

I can give examples about health conditions. Trans women are still at risk of having prostate cancer. If we do not properly understand the relationship between sex and gender identity, we will not be able to analyse whether trans women and cisgender men are more or less likely to have such conditions. Trans men probably still have a higher risk of contracting breast cancer than cisgender men, because they are biologically women. If we do not collect that level of information, we cannot properly understand what risks certain groups in the population face.

There is a concern that people will not answer the sex question, so the question is whether we should have another category. Perhaps we should have a public campaign around why it is important to ask the sex question. I do not know whether many people from the LGBTI community are aware that if they do not answer the sex question, they will be assigned a sex through imputation. If their objection to self-defining as male or female is problematic, their objection to being assigned a sex that they have not decided on would probably be greater.

09:30  

I have a couple of points to make. The census is clearly the most comprehensive data source that we have about populations every 10 years, but committee members should be aware that the census is not without its issues. Obviously, we are moving away from face-to-face collection of the data, but we are not sure what impact that will have on response rates and the accuracy of responses. We know that there are budgetary and other pressures to move the census further away from that, and perhaps even towards a sampling approach, all of which would reduce the quality of the data.

Notwithstanding that the census is the best source that we have, we do not have 100 per cent of the population responding. We therefore already have statistical uncertainty around some aspects of the census. There have been census years that had huge problems in that regard. Famously, in 1991, around the time of the poll tax dispute, there was a very large non-response because people were fearful of being caught up in the dispute. As a result, a large number of people had to be imputed, as Susan McVie indicated, into the data set in order to balance what we thought was the missing population in 1991.

There are problems with the census as it is, so we should steer clear of thinking that the census is accurate down to 0.001 per cent in every parameter that we collect.

A point was made in evidence that we took last week by an individual who said:

“Of course, the sex question is massively important for things such as health planning, but sex is only a proxy for making decisions about sex-specific services ... for example, not all females need cervical screening, because they might have had a hysterectomy. We cannot tell whether someone will automatically need cervical screening just by knowing that they are female.—[Official Report, Culture, Tourism, Europe and External Affairs Committee, 6 December 2018; c 24.]

That person made that point to support their view; it would be interesting to hear your views on it.

We have heard from, for example, Professor McVie, that the data is used more widely than simply for health purposes. However, even if we just take the health purposes and go back to the mandatory sex question referring to sex at birth, or the sex noted on the birth certificate, in order to deal with the legal sex definition, I would have thought, as a woman, that there would be many other potential health implications for those who tick the box to indicate that they are female. Whether someone has had a hysterectomy or not, there will be many other health issues. Am I wrong in that thinking?

Most individual healthcare services will be run through the CHI data set, which is the collation of all a person’s health records, including general practitioner records, prescription records and hospital admissions. That is the system that is used for identifying need for screening services. In relation to the trans community, the variety of healthcare needs within that very broad spectrum will vary widely, so the best way of identifying needs is not through the census; the best way of identifying those needs is through existing health records.

A woman having had a hysterectomy is only one element of her health history, but surely there are many other issues pertaining to someone of the female sex beyond hysterectomy and cervical screening. The argument about that was not compelling, to me. I want to hear what the data statisticians feel about that.

The answer goes back to whether we are thinking about individual need or population-level need—the two cannot be conflated. A woman’s needs depend on a wide variety of characteristics.

Does that go beyond whether one requires cervical screening or has had a hysterectomy? There may be other issues. Professor McVie mentioned breast cancer.

That is true, but none of those would be identified through the census, and nor would people’s needs. Those would be identified through their health records.

I have one last question, and others will probably want to return to this territory. I am interested in the process of the census. Leaving to one side the matter of face-to-face collection, at the moment the census is conducted per household. Concerns have been raised about people’s privacy and about questions that they might feel are intrusive—notwithstanding that it is a confidential process. What are the reasons for the census being carried out on a household basis, and is there now an argument to carry it out on an individual basis, given that the questions are becoming more personal, at least in people’s perceptions?

The facility to give people individual forms is being explored. In the coming census in 2021, people can apply for an individual form, so that they do not have to provide their data as part of a household. That is part of the debate.

However, there are probably other questions on privacy. I am not a historian of the census, but if we go back to 1901, that was a world in which the paterfamilias filled in the family form. We have moved on a great deal since then, as you can see in the planning.

Susan McVie mentioned the publicity and public information for the census in 2021. One of the things that will be very important will be that we make the processes clear to people who, for whatever reason, do not wish to have aspects of their data reported on their behalf. NRS says that it will test further how it will run that part of the census. It is important to note that we are changing that.

That is good news.

That is an important point. It was not clear from the briefing document to what extent piloting of various sensitive questions had been done, taking account of how people in different circumstances would complete them. That has been a big issue in household surveys: the national sexual attitudes and lifestyle survey has sampled individuals within households precisely in order to avoid that. It is certainly the case that there will be many circumstances in which it could be especially difficult to answer such questions. It might also be difficult not to be part of a household response. This is a key point—by not being part of that response, a person is disclosing. The matter needs really careful consideration. It will probably affect only a small minority of people, but it is important to get it right.

I will make a small point on that. Everything that has been said is true, but one of the risks of moving towards individual responses relates to the response rate. It is a balancing act. The classic case is trying to capture the teenagers who are not around when you are collecting the census data, so we just fail to get them. That is a huge problem for voluntary surveys, but it is also a problem for the census. It is just a lesser problem because the census is obligatory.

The more barriers to collecting the data that are put in place, the poorer its quality will be. There is a balance to be struck in respect of whether to get the best available data from one member of the household, or to collect individual-level data in the full knowledge that the response rate will be lower.

I can see that that would be a balancing act, but people’s right to privacy is fundamental. Professor McVie has indicated that there could be a wide-ranging information and take-up campaign to encourage people to complete the census individually.

That would be part of the balancing act. Cost is another factor.

Kenneth Gibson has a supplementary question.

I am not wholly in agreement with Annabelle Ewing on this particular issue. As Dr McCartney said, there is a real issue about missing out large numbers of people. In the committee, last week and this week, we have been discussing the sex question, but it is irrelevant if we do not get hold of the person in the first place. The number 1 priority has to be an accurate population census, and other things are secondary to that. What do the panel think about that?

It is important to note that the proposal is still for a household form—the census is not being moved on to a wholly individual footing. This might be an interesting point to pursue with the Government, but my understanding is that, although it will be a household survey, people can request an individual form. In that case, the question that arises—and which you are raising—is: how far will that create the problems of loss that Gerry McCartney has highlighted?

In order to come out of the household survey, a person must request an individual form, which will mean that 5 million separate forms will not have to be sent out. As Gerry McCartney has said, this is all about striking a balance, and I presume that NRS is hoping that this balancing act will produce the right combination of coverage and protection for those people who are uncomfortable with making a household return.

People can request their own form if they want to.

That is my understanding.

Good morning. You have talked about ensuring that we get accurate information and data. That is vital because organisations, especially in the public sector, need to think about how they provide services on that basis. Is there a good and meaningful understanding of the terminology around gender identity that can capture that kind of meaningful data? Is there such an understanding of how public authorities can use the data that is captured to ensure that they manage their equalities duties effectively and provide the services that are required?

I am not talking about the census setting as such, but there have been many data collections of this kind and, obviously, the piloting that Ipsos MORI does is carried out. For various age and cultural groups, there might need to be a lot of explanation—perhaps more than would be needed if one were to want to justify biological or legal sex. Nevertheless, there are many studies that have provided good-quality data, and they are a great resource to build on.

For me, the issue is ensuring that we have transparency and clarity with regard to the question. It is all about looking at the users of the census and the information that they need. For example, do they need to know that someone self-identifies as trans, which would be a simple yes or no question, or do they need more detail on, say, whether someone self-identifies as a man even though they are a woman? There are all sorts of other gender identities such as gender neutral and gender fluid. The issue is the amount of granularity that users need. Of course, that consideration has to be balanced with the amount of personal information that people from those communities really want to give in a census.

It is all about the benefits of the census, which are that it enables us to see our population’s broad characteristics, to plan and target resources and to do fantastic research. Scotland is at the forefront of some amazing research that is based on administrative data, but we have to identify the benefits of what we are doing and the level of information that users need to benefit the population, and go no further than that. That would be my advice.

The level of data that you will get—and will want to get—from the census is fairly limited, but there will also be things such as the next sexual attitudes and lifestyles survey, which will allow you to make well-founded inferences about those populations and the distributions within them. The census is part of the picture, but not the whole of it.

I think you have exactly identified the issue. The census is one of the things that can be used to understand and support individuals across the spectrum in making an identification and seeing how they fit into the process. In turn, organisations can then look at how they might fit around them in order to provide support. If that happens, individuals will have the confidence to provide the information and make sure that what you get is correct.

Such an approach will help to identify many things, but there will always be individuals who are fearful of giving that information because they think that it might be misconstrued or looked on differently. How will we get everyone to do this? After all, without completely accurate information, we will have only a snapshot.

09:45  

You will not, but you will get very useful information. When the first sexual attitudes and lifestyles survey came out, people said that we would not get any useful information and that people would make it up, but that is clearly not the case. We get good, useful information on issues on which, in some cases, there may be no final answer.

If we are thinking about small population groups and the utility of the data, one of the really important uses of the data is to look at different questions across the census. For example, that can tell us whether people with particular characteristics are more or less likely to be in particular occupations. One thing that we argued for strongly at the previous census was that it should collect income data to enable us to understand whether there were differences according to people’s incomes, but that suggestion did not make it to the final level. Information about the intersectional aspects of the population is missing from the census data. It is information that we can only really get from a census where there is a very large sample size that can be broken down by different characteristics, including protected characteristics, and socioeconomic factors.

I want to explore a couple of issues. Question 2 in the 2011 census asked whether the respondent’s sex was male or female. Is it your understanding that people who completed that census answered that question based on their understanding of their current gender identity or do you think they perceived it as a question about their legal status or biological sex?

As Susan McVie said, one of the problems is that we cannot tell how people read that question. The best evidence that we have is probably in “Scotland’s Census 2021—Sex and Gender Identity Topic Report”, which contains an interesting and quite rich data set of interviews with trans people who had filled in a pilot set of questions, including a binary sex question. Those questions were compared with others, and it was interesting to read that there was a very mixed reaction in that group to the meaning of a question that was pretty much like the one in the 2011 census. We cannot tell whether this is the case, but there is some indication in the topic report that at least some people who identify as trans read that question as a biological sex question.

I was very taken with Susan McVie’s argument that, if we have a really good and clear set of questions in the census, we will be able to backtrack and look at the previous data. However, until we do that, it will be difficult to judge how the question has been read. For most of the population, it will be a very straightforward ticking of a box, but for the group that we are interested in, it is unknowable what proportion will read it in one way and what proportion will read it in another.

Surely the question at the core of this is what the purpose of the census is. What sort of data should we collect? The previous census asked all sorts of weird and wonderful questions about how people travelled to work and whether they were looking for a job. This is about working out how important sex data, gender identity data and sexuality data are. Do you have views on what data should be collected that will dictate what sort of questions we need to ask people?

If we start with the Equality Act 2010, we should be collecting data on sex. Whether we define that as biological sex or legal sex is essentially a matter for debate and will not make that much difference to the numbers.

We then have sexual orientation. There is already a good set of well-tested questions on sexual orientation, so that is not particularly problematic.

We then have gender reassignment, which is a protected characteristic. I do not think that what the 2010 act means by gender reassignment is entirely clear. If we look in detail at the content of the act, the description of gender reassignment and some of the examples that are given are a bit blurry. However, a number of surveys have considered issues around gender and gender identity, and I think we should be going to the tried and tested surveys that have identified good questions that have had cognitive testing to ensure that people understand what they mean and that they are relevant and valid in relation to those individuals.

I am not going to give you a set of questions, but I can say that we are not starting from scratch here. Jackie Cassell can probably tell you more.

I agree with all of that. With regard to the trans question, it will be important to think about what that might mean in terms of the kind of dissonance from either biological or legal sex. Clearly, there is a strong sense, which is accepted in the plans, that there are people for whom that dissonance is problematic. It is not at all clear to me who will answer yes to that trans question. We need to think about what we need to know about the range of people who may or may not choose to say that they belong to the trans category beyond gender reassignment. That is a difficult question and it has not been sorted out yet.

If you want to have a population statistic for trans people, you should ask a yes/no question about trans. You should possibly also give people the opportunity to say how they would self-define their gender identity. Some people would prefer not to do that, but we would not be sitting round this table if there was not a demand from the LGBTQI community for the census to better reflect the characteristics of our population. The issue is about giving people the opportunity to self-identify in the way that they wish, while not imposing a set of questions on anyone. At any rate, the question should not be mandatory.

The question about gender reassignment is a more tricky one, because it is a protected characteristic, which means that we should be collecting that data, possibly through a mandatory question.

My follow-up question is about which questions should be mandatory. It is important that the Government collects the data that we are talking about. It seems to me that there is still debate about whether the collection of data about both legal and biological sex would be useful because, as others have mentioned, there are medical benefits—with regard to, for instance, the diagnosis of certain conditions, as has been discussed—to knowing someone’s biological sex as opposed to how they are currently defined in law.

Which questions involve data that we must know and which are ones that people should be allowed to answer in their own way, regardless of what the question is? I do not think that we need to define that question at this point. Do you have a view on whether the method of data collection will change next time round? Do you think that, if people did not have to give face-to-face answers on the doorstep, they might be more honest in their answers? Might there be disproportionate levels of response from certain communities with regard to the census?

One reason why I strongly support voluntary responses to the questions is that it is clear that, as the guidance discusses, not all of them are meaningful to all people.

An example concerns the history of the collection of sexual orientation data in sexual health clinics. Many years ago, if someone had gonorrhoea, syphilis or whatever, we would collect data on how it had been acquired. At that point, we did not ask people what their sexual orientation was, because there are many people who would describe their sexual orientation as heterosexual but may well have same-sex contact. Sexual orientation is a useful construct, but it is not something that everybody feels is applicable to them in the same way, and that is likely to be true for trans, too.

Where something is not universally felt to be a category that usefully applies, it is not clear that you could or should make it the subject of a compulsory question, quite apart from all the wider issues about privacy.

The question was broader, but can I clarify that you are not suggesting that the sex question should be voluntary?

I do not think that the sex question should be voluntary. The issue is the robust data that we want to have by asking the question. The sex question should not be voluntary, but the sexual orientation and trans questions should be voluntary, for various reasons.

“Biological sex” and “legal sex” are somewhat contested terms as to which is which. I would characterise the issue in this way: are we looking at someone’s original birth certificate sex or their current birth certificate sex? There is no dispute about it being one of the two that comes up as the legal definition of sex; we are looking at one thing or the other. The main issue about the sex question is which of those two it should be.

Tim Hopkins has made an important point, and one to which the committee needs to give careful thought, about privacy rights under article 8 of the European convention on human rights and how far that might bite on the sex question on the census. The Scottish Government does not talk about that in the policy memorandum. My understanding is that we are talking about gender recognition certificate holders who have changed their birth certificate. There is a substantial point, which will be worth teasing out with the Government, about whether the Government thinks that legal privacy rights at that level kick in for that small group of people.

I want to put on the record that, as far as I can see as a data user, the decision could go either way without affecting the data, given that the decision to collect the current birth certificate sex rather than the original birth certificate sex affects such a small number of people. We know a bit about those people because we have a register, so we know their ages and so on; they are not an unknown group. The committee does not need to worry about the data quality impact of the Parliament’s decision on which of the two routes to go for.

I would be interested to know whether the other witnesses share that view.

Professor McVie is nodding.

Yes. If I had to make the choice, I would go for legal sex, which means biological or whatever is on the birth certificate. There is a wider group of people for whom their gender identity is more fluid and less clear, and that is why it is difficult to have a set of questions that fits everyone. “Trans” is used as an umbrella term to describe a community, but many people in the trans community do not necessarily feel that they are the same as other members of the community.

Some will have had surgery and others will not have done so.

Yes. People might be at various stages of medical treatment. Some people might decide that they do not require any medical intervention.

Jamie, have you finished asking your questions?

Will members of the panel respond to my final question, which was about whether the way we collect the data—digitally, through the post or whatever—will affect the data? It might be a question for Mr McCartney as well. Will our asking the new questions that we are talking about encourage certain communities to answer them such that the data set will be disproportionate compared with the response that we would normally get under a different collection method? I am sorry—that was a convoluted way of explaining what I was trying to say.

There are two points there. First, will changing the mode of delivery change the response to the survey? Secondly, will asking a new and quite sensitive question affect the likelihood of people answering it, and will that be influenced by the mode of delivery?

It is quite a complicated set of circumstances because two things are being changed at once. Testing the effect of the changed mode of delivery on the response to the question will be difficult because we will have nothing to compare it with. What we can compare, I suppose, is the levels of response of people who complete the traditional paper questionnaire and people who complete the census using the electronic form. In survey design, we have tested the asking of sensitive questions using non face-to-face methods, and we know that that approach tends to produce a greater response.

For example, in the Scottish crime survey—I think it was back in 2004—we tested a telephone survey to collect data on victimisation. It was a disaster, because people who had not been victims of crime just said, “That’s not appropriate to me” and put the phone down, whereas people who wanted to talk about crime participated. If I remember correctly, we had something like a 150,000 per cent increase in responses to questions about sexual assault because people are much more likely to respond to such questions if they are not being asked them directly, face to face. There are other examples where electronic means make people more likely to feel comfortable about responding.

10:00  

On the overall change in the mode of response, Gerry McCartney said that it is a risk to change from face-to-face to electronic delivery. A percentage of the population will always complete a survey—it does not matter what form it comes to them in—and a percentage probably never will. It is about the people in the middle, and the question is to what extent we have to do more work to persuade them to participate.

Scotland is not the only country that has a census. Do you have any international comparators for the types of question that should or could be asked? I assume that a similar debate will be taking place in other countries.

I am sorry, but I do not know enough about the censuses in other countries to answer that question.

I know that NRS has been working quite closely with the other agencies across the United Kingdom. They have been working together and sharing the research costs of investigating different questions. I suspect that the questions across the censuses in the UK will be quite similar, not only for the purposes of comparison but because the process has been quite similar. I am not sure what is happening beyond the UK.

There are some very large-scale demographic and family surveys around the world, some of which ask fairly detailed questions because there is little other data. I think that the census here needs to sit within the context of very good health data, particularly in Scotland.

In a supplementary submission to the robust evidence that she gave at last week’s meeting, Professor Rosa Freedman wrote:

“The law clearly sets out ... that sex is biological, and that transsexualism (what we would now term transgender) is psychological.”

This morning, we received a submission from Professor Richard Byng, Professor Susan Bewley, Dr Damian Clifford and Dr Margaret McCartney, who said collectively:

“There is little supporting evidence for a genetic or anatomical brain basis for being born in the wrong body, yet this idea now has currency with the public, and it appears they believe it is medically endorsed ... Self-identification could lead to a neglect of the proper, formal exploration of the wider reasons a person may want to transition; these are often unconscious and need time to emerge.”

They go on to say:

“We believe usual standards of evidence should apply (based on the National Institute of Excellence in Health and Social Care) so that interventions improve mortality or quality of life.”

Can you comment on the views in those submissions? How do you feel about them?

In terms of their relevance to the bill—

The submissions are relevant to the bill, because they are saying that, if we do not get proper, accurate information about the individuals concerned, we may make the wrong decisions in relation to interventions.

You said—and Professor Jackie Cassell agreed—that the numbers are small, so you do not believe that the issue would have any impact on how we look at interventions. However, you also said that the number of people who are reporting as transgender is increasing rapidly. I wonder whether, in fact, there is something in what the professors and doctors are saying in those submissions.

Rosa Freedman was particularly interested in the legal definition of sex. That takes us back to the question of whether you are looking at birth certificates, and, if so, which ones. That is a relatively narrow point, whereas there is a broader question about trans identity. We keep coming back to the same point, which is that it is a varied group. We need to get the right bank of questions, including a question about a person’s unambiguous sex status—whether it is attested by one type of birth certificate or another—and, on top of that, questions that measure all the other dimensions of how people identify themselves.

That bank of questions will give the sort of information that is wanted and needed. It is about the layers of information that we collect and not neglecting any one of those layers, because we are dealing with a fast-moving and shifting issue. We cannot even be sure how the language and terminology that we are currently using will play out by 2021. When you put that contribution together with Professor Freedman’s, you can see that you need to be very clear about separating the different types of information that you are collecting. That goes back to the point, which we have discussed a few times, about the clarity of the sex question and the nature of the supplementary questions that will do the best job for this particular community.

That is within the constraints of a census that is not designed to give a comprehensive account of people in detail.

One of the comments that you have relayed to us relates to the evidence base for interventions. I reiterate that the census data will not be used to plan services for individual people; the best source of that information remains within the health service. In a sense, that is an irrelevant comment in relation to the census, because, even if we get the questions perfect on trans status, biological sex and all the different aspects that have been discussed this morning, the data will tell you very little about what individuals need from health services.

You said that earlier, and everyone has taken that on board. I certainly have. I am just trying to give you the perspective of those individuals, who feel that overall medical services may not be designed appropriately if we are not asking the right questions in the census. They are looking for questions that are sex specific rather than to do with gender identification, because they feel that, if you design services on a collective basis, not an individual basis, you might not get it quite right.

The more clarity that we can get in the set of questions and the more clearly we can interpret the data, the better. However, that will only ever be part of the battery of evidence that is available to us.

I agree with Gerry McCartney. The more clarity we have, the better. However, the census is not a survey, so there are limitations on what you can include in terms of people’s background characteristics. The census is intended to be a broad description of the characteristics of the population, which we can then link to other data sets such as the health data sets that Gerry McCartney talked about. The sex question is only one of a number of questions that are used to link those data sets, but having sex in the data is very, very important.

The other supplementary information that we collect around gender identity is important for a broad range of reasons, not just for looking at health service patterns. It would not be used to plan health services, but it would be used to identify, for example, whether people from the trans community take up services to the extent that people from the cis community do or whether people are discriminated against within certain services, such as the criminal justice system. By linking together lots of administrative data sets, we can test all those things.

I agree with a lot of what the panel members are saying. Is it possible that there could be an opportunity to identify where there are clusters? For example, people with certain characteristics may be identified as being in a certain geographic location, so that in one area—let us say Edinburgh—there may be a requirement for specific services, whereas in other areas there might not be.

That is true to an extent. It is easier to imagine that in relation to age. If you have a more elderly population in your constituency, you will know that dementia services need to be more advanced in one area than in another. The same may be true in relation to other characteristics.

Let us return to the important point that Lucy Hunter Blackburn raised about Tim Hopkins’s argument that it is important for the Scottish Government to bear in mind the privacy of people with a gender recognition certificate. Wherever the lawyers seem to be in the debate, there is an understanding that a GRC confers legal sex, but Professor McVie suggested that the issue would not affect the data much. I believe that only 5,000 people across the UK have a GRC, which is a small number, so I understand that point. However, you will be aware that there are moves to change the way in which people obtain a GRC. It has been suggested that, in Scotland or in the UK as a whole, there may be a means of self-identification. Therefore, by the time that we have the census in 2021, people might well be able to self-identify and get a GRC.

The late submission from the clinicians Byng, Bewley, Clifford and McCartney, which has been referred to, says:

“The number of individuals requesting medical assistance for gender uncertainty or dysphoria is rising and the demographic trend is rapidly changing.”

If, by 2021, many more people can self-identify and obtain a GRC, will that affect the data?

Again, it comes down to being clear about what we want to measure. It will affect the data to an extent, because, if legal status is broadened out in that way, we will still be conflating two things: biological sex and legal sex status. From a research point of view, if there are registers of people who have gone through the process, we could connect that data to the census and control for it in doing research. The problem will be if people self-define and the status is much broader and we do not have any measure of how many people are in that community.

Whatever the process is, as long as it is understood, it can be taken into account, although perhaps not perfectly. However, if we do not know what the process is—which is likely to be the case if one simply adds a trans question—that is fine, but we will not know what that represents. If we know what the process is, it will not matter to an extent, because we will know what we are measuring. You might then make different choices, next time round, about what else you need.

Will we know what we are measuring? If somebody had a GRC and they just ticked the male or female box, how would we know that they had a GRC?

We would know at a population level. That goes back to the point that the census is not used to deliver services at an individual level. If we know how many people have gone through the process, their age characteristics and what that process consists of—

But we would not be asking about that in the census.

No, but we would know that a certain number of people had gone through that process, which would allow us to deal with the data in slightly different ways. It is when we do not know how people came to have that characteristic that interpretation of the data becomes problematic.

There is a difference between where we were in 2011, when we had no purchase at all on how people departed from biological sex—we have nothing that we can check that against, unless we can work back from a later census—and a situation in which we know that X thousand people have a GRC, so we know that a certain number of people will say that their birth certificate now says that they are male and a certain number will say that they are female, which allows us to make a good-quality estimation of the impact on the total data of changes in birth certificates.

In using the census, we often bring characteristics together, which is where it gets a bit more problematic or complicated. I am interested in the relationship between education level, earnings and sex. We can clearly measure how that relationship varies over time. If there is a concentration of people changing their birth certificate in a particular subpopulation, we have to start imputing—to use the word that Susan McVie used—quite hard. We have to start guessing backwards a bit by individual case.

That is not ideal, but it is better than pure self-identification, which has no reference point outside itself. At the moment, we can say with real confidence that the scale of GRC holding is so small that we would not worry about trying to impute anything. If a legal change was brought into force in time for the 2021 census—if people follow the process that has been proposed, it would be very quick—it is feasible that thousands of people would take advantage of that to change their birth certificate before the census, which could have an impact in 2021. However, I would worry much less about that than about staying with what was used in 2011, the effect of which is much less manageable or able to be estimated.

10:15  

It is about known unknowns and unknown unknowns. People are registered somewhere. If the definition is their legal sex and they are registered somewhere, that is a known unknown. We do not know from the census whether they are on the register, but we can know that information from elsewhere. We can link those data sets together. We could link the census data to the register data, if that was made available for linkage, which is happening increasingly. We now have the vast majority of the health data, crime data and education data, and we hope to have Department for Work and Pensions data shortly. Increasingly, we are linking all the public sector data so that we have a better understanding of how everything links together, which is seen as standard in the Nordic countries.

The problem arises when we have unknown unknowns. For example, the trans community is not defined in any way, so we do not know who belongs to it. We have a vague and ambiguous question in the census and we do not know the extent of that population from elsewhere, so we have no way of estimating the scale of any problem of bias, discrimination or inequality.

You seem to be as one on that point. However, the explanatory notes for the bill say that the Scottish Government already conflates sex and gender. Furthermore, although we have not had many submissions from public authorities, from those we received it was clear that at least one authority did not understand the protected characteristics or the difference between sex and gender. Even the equality impact assessment for the bill conflates sex and gender reassignment. There is a lack of clarity on the issue right across the Government and public authorities, which you are saying is an unknown and therefore a problem for the gathering of data.

I believe that Professor McVie is on the board for official statistics. Are you concerned about that creep in attitudes in relation to the gathering of statistics?

Yes, absolutely. We know that many administrative data sets do not necessarily use the same definitions, which is why it is important to get it right in the census. The census could be an important model for all our other public sector data sets, which we should harmonise on those questions. If we do not and those things are conflated, there are issues with the Equality Act 2010.

Over time, there has been a fudge, and this is the point at which to do something about it. Through its work on the census, NRS could shine a light for all organisations. Having said that, it is important to recognise that many administrative data sets that are used by public sector organisations are not collected as measures of the population but as management tools that enable the organisations to do a good job. It is not always important that they make a clear distinction between sex and gender identity, because their purpose is to deliver a service to individuals. However, when people are talking about sex and gender identity, they should be clear that those are two very different things. As a society, we have not been good at defining them. People have had a problem with using the word “sex” to define biological sex, because it is connected to forms of behaviour. Therefore, as a society, we have been lazy and have tended to use the word “gender” when we mean “sex”.

We see that in the documents all the time. They switch from one word to the other.

The committee has given the bill a fair level of scrutiny. What do you think about how NRS consulted on this and the previous census, the guidance for which many panel members have suggested was problematic? I was struck by Susan McVie saying that, as a body, you were not consulted on the 2011 census. There seems to be a feeling that the 2011 question on sex was included with little discussion—is that correct?

I was not part of the group that was or was not consulted. My point about 2011 was about the design of the question.

Who did NRS discuss the design of the question with? Is its consultation process broad enough to collect sufficient views before it makes decisions on those areas?

One of the issues with the 2011 census is that we do not know what process went on in the construction of the guidance—the issue is not the question but the guidance on it. Nothing has been said, and I do not know what was done. Perhaps the committee could explore with NRS what the process was in 2011 that led it to make quite a major change in the conceptualising of sex. It does not seem to have been subject to parliamentary or any other scrutiny outside NRS.

On the current process, the topic report is a fascinating read and has lots of great information about the cognitive and quantitative testing of questions. However, it left me, as a reader, with a lot of questions about how, at the decision stage, one decision was taken and another was not. On page 3 of the topic report, there is a strong statement about the importance of sex as a marker in the Equality Act 2010 and elsewhere, but, almost immediately, it says that NRS wants to interpret sex as self-identification. There is a jump from one to the other that is not explained in the document or the policy memorandum.

I had similar questions as I went through the topic report. For example, if it was strongly felt that people must be given a chance to not provide their sex details—as it was too distressing for some respondents to provide their birth certificate sex—why not offer a prefer-not-to-say answer? Why move to a third sex option? One of the trans respondents whose comments were taken up said that there could be a non-response. Reading the process behind it, I struggled a bit to understand why, at various forks in the road, one fork was taken and not another.

Underlying that is the question of why NRS seems to be taking quite a strong view in principle that sex should be regarded as a self-identification issue, and I did not find a clear explanation for that. How far that can be explained by who NRS spoke to or by the process is something that the committee would need to explore directly with NRS.

Following on from that, if NRS was to change its approach to working with a wider set of people, would you be willing to work with it?

Yes.

Yes, as part of the user community.

Okay. We will pass that message on.

As there are no further questions, I thank our witnesses for coming along to give evidence. It has been very helpful.

10:23 Meeting continued in private until 10:48.