United Kingdom Census 2011 – Summary and Analysis

Earlier this year I wrote about the controversy around the question ‘What is your sex?’ in the 2011 United Kingdom census. In that article I established that the question of ‘sex’ was intended to record how the respondent subjectively saw their identity and that the Office for National Statistics and the Census Customer Services were advising transgender individuals to choose the binary option (male or female) that most closely reflected their self-identity, rather than their ‘biological’ or legal status.

I also included my reply to the Office for National Statistics asking for clarification as to how those who identify outside of the gender binary and would not be able to choose either binary option should respond to the question, and whether those answers would be reflected in the census statistics in any way. Read the previous article here.

Sadly I did not receive a response to my questions, but other non-binary gender activists received advice from the Census Customer Services telephone line advising them to enter both male and female if they felt that this was most accurate. They were informed that they would not be prosecuted for failing to answer this legally mandatory question if leaving the question blank, ticking both boxes or writing in a different answer was a genuine attempt to answer the question accurately.

However, at no point did the Office for National Statistics indicate that such non-binary answers would actually be reflected in the eventual census statistics. In fact non-gender activist Christie Elan-Cane‘s 2008 census public consultation period call to add a third ‘non-gender specific’ answer to the ‘sex’ question received a response that showed no intention of recording non-binary genders, and I myself was advised in 2001 that my non-binary answer in that year’s census would not be recorded in the statistics. There was also direct evidence that the answers of people in same sex marriages and multi-partner relationships were being treated as errors and ‘statistically resolved’ in the census statistics.

It seemed that non-binary and genderqueer people talking to Census Customer Services were being given false hope that their answers would be meaningfully recorded in any way.

As such, in May I made a Freedom of Information request using the excellent online service WhatDoTheyKnow.com. In this request I asked the following questions:

Could you please explain:

1a) How is the ‘sex’ question used in census statistics? What is an
answer of ‘male’ or ‘female’ taken to mean?

1b) How the ONS compensates for the inaccuracies/ambiguity
introduced by conflating the separate concepts of sex, social
gender, legal gender and gender identity into one binary question?

2a) Does the census system accept answers for this question other
than responses of only ‘male’ or ‘female’?

2b) Will the figures be made available for the number of people who
answered census question 2 to indicate they are:

i) Both male and female
ii) Neither male nor female
iii) Some other sex/gender, indicated by adding an additional box
or writing an answer in the space around the question
iv) Abstaining from answering the question, indicated by writing
this in the space around the question or by crossing out or
otherwise spoiling the question

2c) Are such figures available for the 1981, 1991 and 2001
censuses? If so, where may I read these?

3a) Will people who indicated that they do not have a single sex
ever have their answer ‘corrected’ or ‘resolved’ to assign them a
single binary sex?

3b) If so, what criteria will be used to assign this sex? How is
this justified?

4) Approximately how many people had their answer for sex
‘corrected’ in the 1981, 1991 and 2001 census statistics for any
reason?

On the 27th of June, Paul Wearn of the Office for National Statistics issued the following response:

1a) Responses to the ***sex*** question, which has been asked since the
first UK Census in 1801, are used, together with age, as the basic
variable, by which the full range of other characteristics, such as
health, employment and unemployment in particular occupations and
industries, education levels, migration, etc are measured. Such
characteristics have always been measured by the sex as reported
subjectively by the respondent. Information on the category of transgender
is not specifically collected in the census since the small numbers
resulting would prevent ONS from disclosing any detailed statistical
information about them, even if a need had been expressed for the census
to collect such information.

1b) For the overwhelming majority of the population ***sex*** and
***gender*** will be the same, and no statistically significant
inaccuracies are introduced by conflating the two. Where someone has
ticked both options or left the question unanswered, a single response
will be created. This is not in any way intended to reflect the true
gender identity of any individual, it is simply done to ensure the
completeness of the final outputs as for every other census question
(except the question on religion which is voluntary). Note that the
scanned image of the original census record, which is stored for 100
years, will retain the original response.

2a) No, the census system does not accept answers to the ***sex***
question other than ***male*** or ***female***.

2b) i) Information on the number of instances of multi-ticking for any
question (including the question on sex) will be recorded, and could be
made available on request subject to the numbers not being disclosive,
once data processing is complete.

2b) ii) Item non-response for all questions (where respondents do not use
any tick or text boxes available) will be published as part of the data
quality report. If any other indication of being neither male or female
was specifically recorded, no figures will be available.

2b) iii-iv) No. This information would not be identified or captured in a
structured way, in the scanning process, although as noted above, the
scanned image will be retained and released in 100 years.

2c) Item non-response results for the 2001 Census are available on the ONS
website (which showed that 0.4% of the population did not answer the
question on sex), but none of the other information requested or item
non-response for 1981 and 1991 Censuses is available.
[1]http://www.statistics.gov.uk/census2001/…

3a) The response as recorded on the questionnaire will not be changed.
However, the data processed from every such record will be edited to
assign the category ***male*** or ***female*** for statistical purposes.

3b) A probabilistic statistical system will assign the sex, based on other
characteristics. This system is called CANCEIS (Canadian Census Edit and
Imputation System), and is used by census offices worldwide.

The system identifies a “donor” record (someone who has answered the
question with a single tick, and has other similar characteristics) and
copies their response. This statistical method is known as
***imputation***.

4) Information about edited records is not available for 1981 and 1991.
ONS has published imputation rates for each variable from 2001 on our
website at
[2]http://www.statistics.gov.uk/census2001/…

Note that the imputation rate will include people who have left the
question blank and those who have ticked both male and female.

Summary

As far as the census is concerned there is no statistically significant difference between ‘sex’ and ‘gender’. The census system does not allow non-binary responses to be recorded for sex, each person recorded must have a single binary answer. Those who answer with both or neither binary options will have their sex ‘created’ for ‘completeness’ by a process called imputation. This is not meant to reflect the person’s actual gender identity, only to make the data ‘consistent’ and complete. As such non-binary gender is not reflected in the census statistics in any way.

The actual answers that people entered on their census forms will be stored, but won’t be made available for another 100 years.

In previous censuses, the numbers of people who entered both or neither option, wrote in their own response or spoiled the question in some way were not recorded. However starting with the 2001 census, both the rate of ‘non-response’ and the rate of ‘multi-ticking’ was recorded and this information was made available in the Edit and Imputation Evaluation Report.

In 2001 approximately 14,000 people intentionally ticked both male and female as their answer to the sex question. A further ~185,000 people failed to tick either box. This totals ~199,000 people who had their answer ‘created’ due to ‘non-response’, which accounts for 0.4% of the census population. Interestingly a further 20,000 people had their single binary answer for sex changed for some reason. Stated reasons for correcting binary sexes include preventing the recording of same sex couples being listed as married or as the parents of a child. These assumptions/requirements are described as ‘hard checks’.

It should be noted that 199,000 was the lowest imputation rate for any question. Questions related to education and employment required as much as 16 to 18% imputation, accounting for as many as 5,400,000 non-responses. Even age and marital status had higher rates at 0.53 and 0.76% respectively.

The criteria for assigning a binary sex when respondents gave non-binary answers or failed to be consistent with expected (highly heteronormative) statistical structures is to identify a ‘donor’ record with similar characteristics for other questions, then take the sex from that household or person. Where possible entire households in a similar local area are used. The choice of sex assigned may involve matching features such as the person’s relationship, partnership, marital or parental status. This may involve judgments based on the sex of ones partner or co-parent, or whether one is a single parent. The report indicates that the validity of imputed sex data was assessed for sample areas by judging the ‘sex’ of the person’s name. By this criteria 75% of imputations were judged to be ‘correct’, with the ‘incorrect’ values more likely to be perceived male names assigned a female sex.

The Office for National Statistics has no plan to change the way it reports ‘non-response’ for the sex question in the 2011 statistics. We will eventually be given a similar non-response rate that will cover the total number of people who omitted or wrote in their answer for sex, and another multi-ticking rate that will reflect how many people ticked both male and female.

Those who answered with a single binary gender but also wrote in a protest at the nature of the question, clarified their gender in more complicated terms or added an unticked ‘other’ box will be recorded as the indicated binary gender. Their protests, comments or elaborations will not be available to the public until the year 2111.

Analysis

The Office for National Statistics has no interest in recording the number of transgender or non-binary gender individuals in the census statistics or recording information such as age, relationship status, health, ethnicity, religion etc related to these individuals. There is no demand for this information from the organisations that use the census statistics and there must be a ‘strong need’ for any information requested. The ONS believes the numbers would be too low to be statistically valid, or that releasing the data would reflect such a small population as to make any information released too specific and identifiable, violating the confidentiality of the respondents.

The non-gender campaigner Christie Elan-Cane advocated for the addition of a third option on the ‘sex’ question during the public consultation period before this census, but was told that people would find the option of identifying as neither binary ‘sex’ too tempting if it were presented to them. Helen Bray of the ONS also informed me that ‘[there is] some concern that such an additional category might encourage some people to simply not reveal their male or female identity, and this could interfere with the demographic analysis we undertake.’

The census statistical systems are designed to ‘resolve’ or erase non-normative genders and relationship structures that do not meet statistical expectations or fit recognised legal structures. The response Zoe O’Connell received from the ONS makes this especially clear. This erasure also includes the answers of binary gender trans* people who indicated that they are in ‘same sex’ marriages, ‘different sex’ civil partnerships or are the parent of a child with a ‘same sex’ co-parent.

In 2001 approximately 14,000 people ticked both male and female and 185,000 people ticked neither box, this accounts for 0.4% of the population. We will eventually be given similar counts of how many people failed to indicate a single binary ‘sex’ or who answered both male and female in the 2011 census. It will be extremely interesting to see if the rate of non-response or the proportion of multi-ticking has risen since 2001 in light of the (albeit limited) campaigns asking non-binary and genderqueer people to tick both answers.

Although “What is your sex?” had the lowest imputation rate for any question, the figures nonetheless indicate that there were almost two hundred thousand answers that were potentially attempting to accurately record a non-binary gender or intersex status, of which the 14,000 multi-ticked answers are highly likely to be intentional. Some of the ‘non-response’ answers counted may have actually indicated a non-binary gender or intersex status by writing this information into the space around the question.

However the ONS has no plans to report figures for the number of people who wrote in, spoiled, amended or clarified their answers on the paper forms. The individual answers will however be stored and made available in 100 years. Knowing these individual figures could be extremely interesting and would help to show how many people felt strongly enough about their non-binary gender to protest being asked for a binary sex on the census. However even with this information, the census data will never be a good indication of the numbers of non-binary people in the United Kingdom due to the intentionally limiting and misleading nature of the question.

We have no way of knowing, until the years 2101 and 2111, how many answers recorded as ‘non-response’ or even as a binary ‘sex’ in fact indicated an unambiguous non-binary answer by writing in this information. We’ll never know how many more people with non-binary genders opted to answer with their assigned or legal sex due to incorrectly believing that was what the census was asking for, due to the legally mandated nature of the question, due to using the online form which did not allow multiple, skipped or written in answers, due to someone else in their household incorrectly answering for them, or out of fear of the ramifications of indicating trans* status on a form that would be seen by their entire household.

Due to the ambiguous nature of the question (asking for ‘sex’ when supporting materials explain that gender identification is required), it is likely that some intersex people with binary gender identities gave non-binary answers to the question in an attempt to accurately record their sex.

Next Steps

Write to the Office for National Statistics requesting that the number of people who wrote in some kind of response extra to the binary options in the question of sex be counted and reported. Ideally this information would be further sub-divided into those who did this while ticking no items, ticking male alone, ticking female alone or ticking both. We would also need to know the number of people who completed the question online and were therefore unable to amend the question or give any kind of non-binary answer. When requesting this information, state that we do not believe that this would be a statistically valid reflection of the numbers of non-binary trans* people in the country, but we do feel that it would give a better reflection of how many felt strongly enough about their gender to clarify their answer or protest the question.

In addition to campaigning about the census now past, if you want the government to legally recognise the existence of non-binary genders and record accurate statistics about our numbers then write to your MP explaining how strongly you feel about this issue and how having your gender ignored and erased impacts your life. Also ask your MP to write to the Minister for Equalities Lynne Featherstone on your behalf to explain how important it is to you that National Statistics surveys and censuses record and reflect non-binary genders and other types of trans* experiences.

Should censuses continue after 2011 (this is currently in doubt), it will be important for more non-binary people to take part in the next public consultation process and advocate for a strong need for non-binary gender to be included in the questions asked.

Visibility and pressure from non-binary people is vital in ensuring that our identities are officially recognised in the future.

---

‎* The asterisk at the end of ‘trans*’ denotes that this is the wider inclusive form of trans that includes all transgender, genderqueer, gender variant and gender non-conforming people regardless of gender identity or expression.