Review of ethnicity harmonised standard: findings from phase 1 of our research
This report summarises the main findings from the first phase of our work to review the current ethnicity harmonised standard.
|Publication date:||9 March 2023|
|Owner:||GSS Harmonisation Team|
|Who this is for:||Users and producers of statistics|
|Type:||Harmonisation standards and guidance|
The Government Statistical Service (GSS) Harmonisation team is based in the Office for National Statistics (ONS). The team supports harmonisation of data across the UK. This involves providing bespoke advice and harmonised standards and guidance about how data on different topic areas are collected and presented. The standards and guidance are designed to improve the consistency, coherence, and comparability of statistics.
Producers of statistics use these harmonised standards as a starting point in their data collection process. By using harmonised standards, they can:
- align with other producers of statistics
- increase the usefulness of their statistics
- meet the Code of Practice for Statistics’ cross cutting theme of coherence
Harmonised standards also allow people to effectively and accurately compare data that has been collected across different datasets. This means we can more easily understand what that data does, and does not, tell us. This ensures that statistics are being used to their full effect for the public good.
Ethnicity is a multifaceted and changing phenomenon. This means various ways of measuring ethnic groups have been used over time. This report presents the first phase of our work to review the current ethnicity harmonised standard. Read the report about the second phase of our work.
- provides contextual information about what ethnicity is and about the current standard
- sets out the reasons for conducting this review, including the approach we took
- summarises the findings from our initial research and engagement — this includes an exploration of issues with the current standard and information about user needs
- sets out our future work and gives associated timelines
As part of this review, we have worked with a variety of government stakeholders and the public to gather feedback on their experience of using the current standard. This included asking about areas where these stakeholders feel the standard could be improved.
We have only partially completed our planned engagement at this stage. This report presents our findings to date. We are publishing our first findings now to provide transparency and to gather feedback on our work conducted so far. We have further research and engagement planned in the coming months.
What we mean by ethnicity
There is no single definition of what makes an ethnic group. People decide for themselves whether they belong to a particular ethnic group and their ethnicity is subjectively meaningful to them personally.
Because the concept ethnicity is multifaceted and constantly changing there are various possible ways of measuring ethnic groups. Many different approaches have been used over time. These include measuring ethnicity using criteria such as common ancestry and elements of culture, identity, language, and physical appearance.
Someone’s ethnicity is not a static concept. Ideas about what makes an ethnic group may change according to the context of social and political attitudes or developments. People are free to choose how they identify and define themselves when they are answering questions about ethnicity, and this might differ if they are asked the same question more than once.
The current harmonised standard
The current ethnicity harmonised standard is based on the 2011 Census questions across the UK. The questions were reviewed and adjusted so they could be used in the:
- 2021 Census for England and Wales – Wales having the addition of Black Welsh and Asian Welsh in the top-level categories
- 2021 Census for Northern Ireland
- 2022 Census for Scotland
This means that the 2021 and 2022 Census questions are, currently, the most up to date way of collecting data about ethnic groups.
We currently recommend you use detailed ethnicity classifications wherever possible. We also recommend you consider using the national identity harmonised standard and religion harmonised standard when asking any questions about ethnicity. This helps people give details about their full cultural identity. You should ask the questions in the following order: national identity, ethnic group, and religion.
We have also worked with the Cabinet Office’s Equality Hub on their standards for ethnicity data, that are currently in draft pending the results of the public consultation on them. The standards bring together best practice and guidance about ethnicity data. They give public sector data producers and data users information about how to better collect, analyse and report on ethnicity data. The standards include guidance and information on:
- the importance of using harmonised standards
- how to aggregate and analyse ethnic groups
- the importance of understanding and reporting on missing ethnicity data
The standards for ethnicity data clearly state that users should use the GSS harmonised standards, or explain clearly why they have not been used, as required by the Code of Practice for Statistics. The Government Equality Hub’s work complements the GSS harmonised standards by providing additional support on ensuring trustworthiness, quality, and value.
Reasons for this review
ONS undertook extensive ethnicity research and public engagement activities as part of development of the 2021 Census ethnicity question. But we know ethnicity is a topic that is evolving. Definitions, terminology, and thinking on the topic are changing and developing. The Census 2021 question development was primarily completed before 2020 and new drivers for change have emerged. Events such as the Black Lives Matter movement have affected how people feel about ethnicity. The Coronavirus (COVID-19) pandemic and the disparities in outcomes and prevalence between ethnic groups have demonstrated the need for more granular ethnicity data.
This review will add to previous work that was completed as part of Census 2021 question development. It will allow us to think about a range of user needs and consider an updated standard to meet these needs.
This review is an opportunity to respond to the Inclusive Data Task Force’s recommendation to ensure concepts are being measured clearly and appropriately. The review will also ensure our ethnicity standard reflects the diversity of the population. Additionally, the review presents activities completed to build trust with those supplying data and those using data to improve inclusion.
To complete this review, we have undertaken four broad areas of research.
Our research activities included:
- reviewing published academic literature and current initiatives, such as research and analysis published by the Race Disparity Unit (RDU)
- analysing queries and feedback received by the Office for National Statistics (ONS) and Government Statistical Service (GSS) from members of the public and organisations about the current ethnicity question
Engagement with people who use or collect ethnicity survey data
Our research activities included:
- commissioning an online survey targeting data collectors and processors across government
- leading a workshop with the Ethnicity and Religion Assurance Panel — this is a group of stakeholders and independent experts brought together by the ONS
- working with the Harmonisation Champions, a network of representatives across government and non-governmental organisations
The GSS Harmonisation team have an ongoing survey of data collectors and processors. It gathers information about other surveys that use harmonised standards and includes information such as data collection modes and capabilities. The survey captures information across our harmonised standards and provides us with general information, such as ability to use free text.
We used information from the survey up to July 2022. This provided a sample of 43 respondents from a range of organisation types, including central government departments, devolved administrations, and the private sector. There was a good balance between data collectors and processors in the sample. 40% of respondents were from devolved nations, which represents a large proportion of the sample. To gather this sample, we:
- identified important stakeholders across UK government, devolved nations, and the private sector
- selected relevant data collectors and data processors from the list of stakeholders to be contacted to fill in the survey
- contacted GSS Harmonisation champions to help identify additional data collectors and data processors that use the harmonised standards
- conducted an implementation review to identify surveys that use harmonised standards
- contacted producers of official statistics that use harmonised standards and additional contacts identified through follow up stakeholder engagement
Engagement with people who use or collect ethnicity administrative data
Our research activities included:
- holding workshops with administrative data collectors about the collection of ethnicity data
- gathering feedback from government departments and the Devolved Administrations
- speaking to people who review and explore data linkage from administrative sources — this helped us increase knowledge about how this data is collected at source
Initial engagement with the public
Our research activities included:
- reviewing queries received by ONS regarding the Census 2021 ethnicity question — this includes comments from members of the public and organisations
- completing an unmoderated card-sort to explore participants’ mental models of their ethnicity
An unmoderated card-sort exercise was undertaken in August 2022 using the Optimal Workshop online platform. Participants were recruited by People for Research, with a prize draw offered as an incentive. This activity aimed to explore the different terms and concepts that people use to describe their ethnicity.
The task presented participants with a list that gave a breakdown of different types of response options that are currently used across government to collect data on an individual’s cultural identity. The response options were written on separate cards. Participants were asked to look at the list and choose as many cards as they liked that they felt reflected their ethnicity.
We created a total of 30 card labels based on breaking down response options and data outputs currently found in the:
- ethnicity harmonised standards for England, Wales, Northern Ireland and Scotland
- religion harmonised standard- we specifically used the data outputs for GB and UK data presentation
- national identity harmonised standards for England, Wales, Northern Ireland and Scotland
- ethnic group questions used within the various UK censuses in 2021 and 2022
We categorised the cards at the start of the activity according to their current use within ethnicity questions as predominantly either:
- external based terms- for example, White, Black, Mixed, or Multiple
- region based terms- for example, Asian, African, Caribbean, or Arab
- country based terms- for example, Chinese, Bangladeshi, Pakistani, Indian, Irish, English, Welsh, Scottish, Northern Irish, Polish, British, or Filipino
- heritage or culture-based terms- for example, Gypsy, Irish Traveller, Roma, Showman, or Showwoman
- religion based terms- for example, Christian, Buddhist, Hindu, Jewish, Muslim, or Sikh
Participants who chose more than one card were also asked to put the cards in the order that shows how strongly they identified with the different terms.
Analysis and findings
The unmoderated card sort had 1425 responses for ethnicity.
The cards that were chosen most often in our sample were:
- White – this card was used 837 times
- British – this card was used 746 times
- English – this card was used 586 times
- Christian – this card was used 213 times
The term Mixed was used 87 times. It was used more often than the term Multiple, which was used 56 times.
95% picked between one and four cards to describe their ethnicity. Participants selected 2.4 cards on average.
Participants who selected “Mixed”, “Christian”, “Multiple”, or “Black” as their first card, used more than the average number of cards to describe their ethnic group.
Participants who selected “Mixed” as their first card, picked the most cards to describe their ethnicity. These participants chose 3.3 cards on average.
Participants who selected “White” as their first card used the least cards. These participants chose 2.2 cards on average.
About a third of participants selected one card only to describe their ethnicity.
Some participants picked options that are not currently available in the harmonised standard, such as Christian, Sikh, Muslim, Hindu, Polish, and Filipino.
A common theme was a missing ethnicity response option. This was predominately European or specific terms describing European countries or regions.
Phase 2 of our research
We have now published the findings from the second phase of our research. As part of this research we have:
- commissioned a survey for ONS interviewers to explore their experiences of asking ‘What is your ethnicity’
- worked with RDU to explore the public perceptions of ethnicity terminology — initial findings will be available at the end of March 2023 from our research partner, Basis Social
We will also be running community and Local Authority engagement sessions to identify user needs for the ethnicity harmonised standard.
Research and engagement findings: identified issues
We identified 3 main high-level issues in our first phase of research.
Issue 1: selecting a response option
The way the question is currently asked makes some people feel marginalised. This is because they do not always identify with the response options and often leads them to request additional tick boxes. For example, tick boxes do not allow people to specifically identify as British, English, Northern Irish, Irish, Welsh, or Scottish unless they first identify as White. This is partly because concepts such as national identity, geography and “race” have been brought together to form these response options.
The public are challenging our current approach of asking people to choose a single option when they are asked about their ethnicity. By allowing the public to choose more than one option, we could provide more inclusive options for someone to self-identify their ethnicity. In our card-sort research, participants selected an average of 2.4 cards to describe their ethnicity. A third of participants only selected one card and 95% of participants selected between one and four cards.
Some participants used different terms and concepts in the card-sort that are not reflected in the current response options on ethnicity. The terms used in these cases were often the same terms used in the religion question commonly used across government.
People with European identities, especially those from ethnic minority groups, have noted difficulties with knowing where to record themselves. This risks accidental mis-categorisation because the “White” category places more emphasis on British identities.
Issue 2: terminology use and presentation
During the development of Census 2021, there was criticism from the public about the use of colour terminology used in the question. Acceptable terminology for collecting ethnicity has changed throughout recent history.
The response options may influence how people view, consider, and understand ethnicity, as well as associated terminology and groupings. We are aware that members of the public can view the question as being preferential to White and British culture, but there is usually a reason behind the design decisions which lead to this perception. For example, ordering of response options is usually based on frequencies of the response options being selected.
Members of the public say that blending or conflating internal identity and external perception factors within the question can make it confusing, difficult to answer, or offensive.
Issue 3: the Mixed category
Our research indicates that the Mixed category receives criticism that it is not comprehensive enough. It can also be seen to prioritise the White part of their Mixed ethnic group due to how the response options are presented. We know from the card-sort that people who select “White” as their primary option tend to use fewer cards on average to describe their ethnicity. People who selected “White” in our card-sort used an average of 2.2. cards. People who select “Mixed” tend to use more cards. People who selected “Mixed” in our card-sort used 3.3 cards on average.
Research and engagement findings: identified user needs
User need 1: ethnicity data which can be compared across multiple dimensions
Stakeholders who use data need to compare their ethnicity data with population-level census data to quality assure and contextualise their data. They also need to be able to compare data with government datasets.
Some stakeholders need to compare ethnicity data across different levels of geography. For example, some data analysts value being able to compare across the four nations, but also at lower levels. The ethnicity harmonised standard gives guidance on how to aggregate data to a comparable UK level, but many stakeholders asked whether we could create a single question to be used across the whole of the UK. This would help overcome the technical or financial limitations that some data collectors encounter. This is especially true of administrative data which is collected for specific purposes. Administrative data is not collected with the main aim of being comparable with other statistics and the cost of providing separate country-specific questions may not be justified.
All stakeholders need to compare ethnicity data over time. Data comparability is supported using consistent question sets and terminology.
Stakeholders who collect data need advice about how to ask the ethnicity questions across different survey modes. Census, which is commonly used by many data collectors, is self-complete. There was little guidance for the many data collectors who are still interviewer led for example when telephone interviewing. The issues were largely about conflation of concepts in the question and group-specific challenges, for example ‘othering’ response options.
User need 2: a harmonised standard which works for stakeholders
Stakeholders need more information about various aspects of outputting ethnicity data. This is because data users have identified that it is unclear what the question is measuring. Output needs included:
- guidance on how response options, in particular the ‘other’ category, should be coded and aggregated, for example when linking administrative data
- guidance on how the ethnicity question fits with other cultural identity questions
- a way to consolidate guidance across government on use of terms
User need 3: aggregate levels of ethnicity data
Stakeholders need to both collect and report ethnicity data at a granular level because it can be desirable for analytical purposes. But small sample sizes in surveys challenge their ability to do this as it can lead to potentially identifiable data. If there is a significant risk of identifiable data ethnic groups are often aggregated and reported using the five high level categories, as used in Census 2021, or binary categories of White and ethnic minority. This is not useful for many stakeholders, including those who produce statistics and government policy makers.
Small sample sizes can be caused by:
- low numbers in the general population
- Incorrect sampling frames
- non-response bias
- respondent burden leading to non-response
- the question being voluntary
Stakeholders reported that multiple response options and write-in options can cause issues for coding and aggregating data, as well as affecting multivariate disaggregation and intersectional analysis. Free text is often possible to collect, but it can be difficult to process the responses because of resource limitations.
Multivariate disaggregation is the process of breaking data down into groups to look at multiple elements of the data at the same time. This allows users to look at the combined influence of these elements on different outcomes. For example, users expressed a need to disaggregate the data by ethnic group and socio-economic status. This would allow users to look at the outcomes of people from the Bangladeshi ethnic group who are on a low income, rather than all people from the Bangladeshi ethnic group.
Users also mentioned the need for intersectional analysis. This is a specific type of multivariate disaggregation which involves breaking the data down into groups that relate to prejudice and equality. This includes topics such as gender, age, sexual orientation and disability status, as well as ethnicity. This type of analysis could be used to look at the outcomes of Black, gay women.
User need 4: change implementation and the need for time
While change is often possible, evidence suggests there will be a lengthy lead-in time for change and at least one year will be needed for change implementation. In addition, there will be potential extra cost, especially for administrative data. Lead-in time is influenced by how often data are collected. For example, a census will have a much longer lead-in time for change than an annual survey because of how often data is collected using this mode.
The timing of a change within the life cycle of a survey or administrative data collection will affect how easy it will be to make the change.
Lead-in time for change is also strongly related to the need for support from stakeholders and partners for the change.
User need 5: support from stakeholders and partners for change
If a change to the data standard is needed, data collectors and processors would need to actively engage with a range of stakeholders and partners to discuss and gain approval for the change. Getting agreement from stakeholders for change will take more time and effort for data users who do not own the data.
It is important to remember that stakeholder engagement is tiered and complex. This must be considered if changes to the ethnicity data standard are implemented. We will need to support and provide evidence that proves the need for change. It will take data collectors and processors significant time and effort to engage with stakeholders to:
- discuss the change
- negotiate how the change will take place
- gain agreement from their stakeholders for the change
User need 6: clearer guidance on what is being measured
Stakeholders were asked about which aspects of ethnicity they were most interested in. They expressed interests in different aspects of ethnicity for different reasons. For some stakeholders, skin colour for visual aspects of discrimination was important. For other stakeholders, religion or national identity were important.
Through our research and engagement activities to date, we have identified that ethnicity is a multifaceted term that often means different things to different people. We also identified that some of the current ethnic group classifications we use in the harmonised standard may require changing.
People who use ethnicity data have multiple needs. They can find it difficult to work with the data because it is complex and measures multiple aspects of ethnicity.
We will continue to investigate needs and issues and we will present these in our second publication.
What happens next
We will continue our work on this topic over the coming months concentrating on:
- the public
- community groups
- people involved with administrative data — this includes people collecting and outputting these data
- civil society organisations
- other interested parties both within and outside of government
Once this work is completed, we will:
- publish a further update in late Spring 2023 on the remaining findings of our continued research and engagement — the update on the findings from the second phase of our research is now available
- identify potential changes to the design of the ethnicity question used within the current harmonised standard and its associated guidance to address the diverse user needs that have been raised by those using the data and the respondents
- undertake cognitive interviewing with members of the public to test any potential changes to the question and its accompanying information
Our second publication includes a new timeline for the development and testing of the harmonised standard in the late Spring 2023 publication.