FOI Ref: FOI/2022/4085
You asked
Please provide details of how inconsistencies between responses to the sex question and the gender identity question in the 2021 England and Wales census will be handled.
For example: if a person records their sex as male in the sex question, and then records that they are a trans man in the gender identity question (and vice-versa).
Please state whether imputation has/is being used to handle such inconsistencies, and if so, provide details of any imputation rules - for example, whether it is the sex reponse or the gender identity response that is subject to change.
We said
Thank you for your request.
In practice, we would not know for certain whether an individual response was inconsistent. In your example, the trans man in question might have had a Gender Recognition Certificate (GRC) and answered the sex question on that basis.
More broadly, imputation is only used in the census in two circumstances:
To estimate the characteristics of those in the population for whom no census response has been submitted.
To estimate specific characteristics in cases where a census response has been submitted but with individual questions omitted. This only applies for mandatory questions, so not for the voluntary questions on religion, sexual orientation and gender identity.
In your example, the sex question already had a response, so there would be no imputation. The responses to both questions would be left as is, feeding into the relevant statistical totals.
On no occasions would a gender identity response be used to impute sex. Where gender identity values are missing, they are not imputed but are treated as valid, and will appear in the "Not stated" (or equivalent) statistical totals.
Our approach to imputation applies a well-established standard which we have used since 1981. More information is available in the "Process and estimate" section of Statistical Design for Census 2021, England and Wales.