Policy divergence in the UK census

Every ten years the census gives us a complete picture of the nation. It allows us to compare different groups of people across the United Kingdom because the same questions are asked, and the information is recorded, in the same way throughout England, Wales, Scotland and Northern Ireland.
Why we have a census: Office for National Statistics

On Sunday 21 March 2021 the decennial census will take place in England, Wales and Northern Ireland. Following an options assessment, Scotland’s census will now take place a year later, on 20 March 2022, due to Covid-related workload pressures cited by the National Records of Scotland (NRS). While described by officials as ‘the right decision for Scotland’, the delay nonetheless breaks with the 2015 statement of agreement between the UK census authorities, which provides that the ‘three censuses should be carried out with reference to the same date in 2021’. It is also notable that ‘Stakeholders were not consulted about the change of census date‘ in Scotland.

The same joint statement also sets out an agreement in relation to the key definitions and classifications used in the census, aimed at securing comparable outputs across the UK.

‘Common definitions and classifications, typically based on international standards, should be agreed, used and published. [This includes a common population base.]

Common topics and questions should be agreed wherever possible, with the intention of making available consistent census outputs across the UK.’

Focusing on Scotland, England and Wales, this blog considers the potential for policy divergence as a result of recent developments in relation to the sex question guidance, and the implications for data quality and comparability.

Sex in the UK census
Across the UK, all census respondents will be asked to state their sex in a mandatory and consistent question that simply asks if a person is male or female.  

Until recently both the Office for National Statistics (ONS) and National Records of Scotland (NRS) also intended to provide additional guidance that advised respondents to answer the sex question based on their self-declared gender identity. However following a series of recent interventions (see below), the ONS is now reviewing this approach. Meanwhile, the NRS remains wedded to guidance based on gender self-identification. In recent correspondence with Professors Lindsay Patterson and Susan McVie at the University of Edinburgh, the Office for Statistics Regulation (OSR) confirmed that NRS’ position that ‘the guidance is finalised… is a policy choice by the NRS’ (and did not, as the NRS had previously implied, reflect the democratic wishes of the Scottish Parliament). In addition to concerns about data quality, this now raises the possibility of policy divergence across the different censuses in relation to the sex question definition and outputs.

Policy capture and how we got here
That the NRS has declared its intention to depart from the standard understanding of sex to incorporate ideas about identity, while the ONS is labouring with its definition even at this late stage is indicative of the conceptual confusion that has been sown around a physical characteristic that is a) easily and accurately recorded at birth for all but a very small minority of people, and b) widely understood as a key determinant of physical, economic and social outcomes.

In The Political Erasure of Sex: Sex and the Census Jones and Mackenzie detail how this remarkable situation came about. Tracing the question development process by ONS and NRS, the analysis shows how the sex variable became corrupted, in part as a result of the unfettered access enjoyed by organisations which advocate the over-writing of sex with gender identity, coupled with a lack of wider consultation, transparency or democratic oversight. As a relevant aside, similar dynamics can be observed in data collection practices across a range of public policy fields, including criminal justice, Higher Education, and health. We have previously highlighted the close relationship that Stonewall appears to enjoy with ONS. A similarly close relationship appears to exist between NRS and the Equality Network in Scotland. The upshot, as Professor Selina Todd observes in her introduction to the Jones and Mackenzie report, is that it is ‘uncertain that the 2021 census (2022 in Scotland) will collect robust, high quality data on sex.’ A recent Freedom of Information response has now shown that the ONS paid over £35,000 between 2009/10 and 2019/20 to Gendered Intelligence, Isle of Wight Pride, Mermaids, Pride Cymru, Stonewall, and Student Pride Ltd for training, attending events, and ‘other goods and services’ including membership of the Stonewall Diversity Champions programme.

Challenges and recent interventions
From late 2017, the politicisation of the UK census increasingly came into public view, prompting concerns about data quality from stakeholders and data users, criticism from feminists concerned about the loss of sex as a meaningful political category, and more recently, challenging questions from oversight bodies.

In December 2019 over eighty quantitative social scientists wrote to the ONS and NRS, expressing concern that the guidance would ‘actively undermine data reliability on a key demographic variable, and damage our ability to both capture and remedy sex-based discrimination’.

In September 2020, as part of its ongoing accreditation assessment of the census the Office for Statistics Regulation (OSR) wrote to the ONS, questioning their assumption that the impact of the guidance on data quality would be negligible, particularly at a sub-group level (for example, when sex is cross-tabulated with age).

‘The assessment team thinks it essential for ONS to consider the concerns raised by users during its further testing and research on the guidance on the sex question, and consider the impact of data quality on the analysis of small sub-groups of the population’.

On 30 October 2020, Sir Bernard Silverman, Chair of the Methodological Assurance Review Panel for the 2021 Census wrote to the ONS National Statistician, stating:

‘Whatever ONS’s view of the aim of the sex question, and whatever the wording of the guidance, the Panel expects to be given a clear assessment of any possible bias or distortive effects, both directly on the question itself, and also on other aspects of the methodology of the census, such as approaches to linkage, undercount and imputation, and future uses of the census’.

The Panel also advised that those raising concerns about the sex question guidance ‘be given written feedback addressing issues they have raised, at an appropriate level of detail and rigour’.

The following week ONS updated its website, stating that its previously published report on guidance for questions on sex, gender identity and sexual orientation related to the 2019 census rehearsal, and that the ‘guidance to be used for Census 2021 is currently under review’.

Policy divergence, data quality and comparability
Taking an overview of the current situation, several observations can be drawn. Firstly, it is worrying that NRS continue to adhere to a position that ignores the data quality concerns raised by relevant academic experts and data users. Although directed at ONS, the recent interventions around data quality by OSR and the Methodological Assurance Review Panel are also relevant to the NRS. A senior NRS official admitted in September 2019 that it was not known what impact similar guidance in 2011 had on the quality of data on sex (Wilson. Col. 6). Moreover, without robust data on the transgender population it is difficult to see how the NRS can be confident in the quality of its data at a sub-group level. That a recent population representative Swedish study reported 6.3% of 22 to 29-years olds wished to be treated as the opposite sex should also ring alarms bells. With just over four months until the population count in England and Wales, it is equally concerning that ONS remains undecided as to how a core demographic variable is defined.

The decision to delay Scotland’s census has already damaged alignment between Scotland and England and Wales. The prospect of policy divergence in relation to the sex question now represents a further risk to our ability to undertake comparative analysis across the United Kingdom. With anticipated programme costs of £800 million in England and Wales (see para. 4.6) and £100 million in Scotland (see para. 9) (pre-Covid estimates) these risks to data quality and comparability carry an exceptionally high price-tag.