Review of the ethnicity harmonised standard: overview of our redesign research to date

There is no single definition of ethnicity. It is a multifaceted and subjective concept. To produce meaningful statistics, it is common for broad ethnic groups to be offered as response categories. People are then asked to decide which broader ethnic group aligns best with how they would identify themselves.

There are various ways of defining ethnic groups. Many different approaches have been used over time. These include measuring ethnicity using criteria such as:

• where you were born
• heritage — culture and ways of living passed on from past generations
• culture — including celebrations, food, values, style, and other elements
• national identity — affiliations and connections with a country or nation
• religion — whether you are practising or brought up in a specific religion
• skin colour
• language

A person’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 can choose how they define themselves when they are answering questions about ethnicity. Their answer might differ if they are asked the same question at different times or in different situations. It may also change if they are asked about their ethnicity for different purposes, or by different people. How someone chooses to identify can also change over their lifetime.

Respondents do not always identify with the current response options, and this has led to requests for additional tick boxes.

One of the key challenges identified in our research is the difficulty some respondents face when selecting a response option under the current harmonised standard. A major contributing factor is the routing logic between Question 1 and Question 2, which restricts the combinations of responses that can be selected. For example, a respondent cannot currently select both “Asian” and “Roma,” even if both reflect their identity.

To address this, we have tested presenting all response options to respondents at once, regardless of their initial selection. This approach allows for greater flexibility and acknowledges the complexity of how individuals define their ethnic group.

We are also aware of the marginalisation some individuals feel when forced to select from one of the many “Other” categories. In the redesigned question, we are testing a single, clearly positioned space for write-in responses, to ensure that those who do not see themselves reflected in the tick boxes still have a meaningful way to self-identify.

Our Discovery phase identified eight distinct aspects that people may use, individually or in combination, to define their ethnic group. Given this diversity, and the self-defined nature of ethnicity, we have chosen not to include prescriptive guidance within the question itself. This respects the personal and fluid ways in which people understand and express their identity.

We are also exploring the use of multiselect functionality, which would allow respondents to select more than one option. This could support more granular and inclusive reporting, particularly for individuals with multi-ethnic backgrounds.

Finally, our Alpha testing revealed that some European respondents perceive the “White” category as being primarily associated with “White British” identities. To address this, we are testing the inclusion of guidance within the response option itself, making it clear that “White” includes European identities as well.

Examples of proposed changes:

Old question: The respondent first selects a high-level ethnic group, and then only the sub-categories related to that group are shown

New question: All ethnicity sub-categories are shown to all respondents, regardless of which high-level ethnic group they select

Old question: 5 categories of “Other” tick boxes

New question: Single “Other” write in

Old question: Select one response

New question: Select all that apply

Acceptable terminology changes over time – our research suggested that combining internal identity and external factors within the question can make it confusing, difficult to answer, or offensive; additionally, the current presentation of response options does not use an alphabetical ordering and can be seen as preferential to White British respondents.

Public acceptability of terminology, particularly colour-based terms, has been a key focus throughout Alpha testing. Our findings show that respondents are strongly influenced by the structure and language of existing ethnicity questions, particularly those used in the UK Census. When asked, “What is your ethnic group?” most respondents defaulted to familiar Census categories, reinforcing the expectation for a question format that aligns with established standards to reduce respondent burden.

Given the concerns raised during Discovery about the use of colour terminology, we tested alternative approaches, including reordering the question structure. While respondents were able to complete the question accurately, this change created challenges for those who identify using response options such as “Asian Other” and “Black Other”, where the expected structure was disrupted.

We also closely monitored the acceptability of the term “Black” and its presentation. While there is evidence of evolving views, the majority of respondents who would have identified as Black in the current harmonised standard, continue to do so in alternative question designs. We will be retaining the current terminology to ensure consistency with the five high level ethnic group categories which will support adoption of the revised question across government and public services. This remains an area of ongoing sensitivity and will continue to be explored in future testing phases.

To address concerns around ethnocentrism, we trialled an alphabetical ordering of response options. This approach was consistently well received across testing and had no negative impact on the respondent experience.

Discovery also highlighted that the current harmonised standard combines internal identity (how individuals see themselves) with external perception (how others may categorise them). In early Alpha, we explored separating these dimensions into distinct questions. However, this approach increased respondent burden and conflicted with expectations shaped by the familiar format of the harmonised standard. It also introduced challenges for maintaining consistency in time series data, which is important for data users tracking trends over time.

Our ongoing work aims to balance clarity, inclusivity, and usability, ensuring that terminology and presentation support accurate self-identification while remaining acceptable and intuitive for respondents.

Examples of proposed changes:

Old question: 5 high-level ethnic group categories

New question: 5 high-level ethnic group categories (unchanged)

Old question: Frequency-based ordering

New question: Alphabetical ordering

Our research suggests the tick boxes within the “Mixed” category are not comprehensive enough and are sometimes understood as prioritising the ‘White’ part of their Mixed ethnic group due to the way the response options are presented.

In addition to changing the response option ordering to alphabetical, we have extended this approach to the presentation of the Mixed categories. Currently, these categories are centred around White ethnicities:

• White and Black Caribbean
• White and Black African
• White and Asian

We are also exploring the inclusion of other relevant ethnic groups, such as Asian and Black. However, the benefits of adding more tick boxes must be balanced against the potential impacts on the ethnicity harmonised standard. These impacts include data comparability over time, the complexity of data aggregation and output, and the burden on respondents.

Examples of proposed changes:

Old question: Four options for Mixed ethnic group

New question: Expanded options for Mixed ethnic groups

Some people understand ethnicity as defined by an individual factor (such as heritage, culture, national identity, religion, language or physical appearance) and other people consider the concept of ethnicity to be influenced by a combination of these factors.

Our research continues to support the eight factors identified during Discovery that individuals may draw on when defining their ethnicity. Respondents generally do not expect a formal definition of ethnicity when answering the question. Most are confident in their ability to self-identify based on their own understanding, even when that understanding differs from others’. This reinforces the importance of maintaining a flexible, self-defined approach to ethnicity data collection.

For interviewer led modes, however, there is a need to provide general guidance that supports respondents without prescribing a definition. Given the personal and varied nature of ethnicity, any guidance must be carefully worded to avoid limiting how individuals interpret the question. As part of our ongoing research, we will explore how best to frame this guidance to support both respondent understanding and interviewer delivery, while maintaining consistency across modes.

This insight underscores the complexity of designing a question that is both inclusive and intuitive, while respecting the diverse ways in which people understand and express their identity.

Examples of proposed changes:

Old question: No guidance on how ethnicity can be defined

New question: Interviewer-led modes to include guidance to support respondents without prescribing a definition

Our research suggests that some people who consider ethnicity as defined by genetics and family history also understand it as fixed or static; this is in contrast to other people’s experiences of how they identify their ethnicity, which can change in different contexts and over their lifetime.

Ethnicity is widely recognised as a self-defined and context dependent concept, and views differ on whether it can or should be treated as a fixed characteristic over time. This issue becomes particularly relevant in longitudinal datasets and administrative records, where individuals may report different ethnicities at different points in time.

In collaboration with colleagues who published the Analytical Learning Points for Ethnicity Data report, we have considered the implications of this variability. Learning Point 3 from that report recommends using an individual’s most common ethnicity across records, rather than the most recent, as this approach aligns more closely with population measures from Census data and supports consistency in analysis.

Exploring public views on the fluidity of ethnicity is methodologically challenging, as it requires identifying individuals who perceive their ethnicity as changeable in the first instance. However, we did test public preferences around how ethnicity should be reported in relation to output granularity. When presented with their granular response option output versus the aggregated output category, such as “Black Caribbean” versus “Black”, participants generally preferred the more granular output, as it was seen as a more accurate reflection of their identity.

These findings reinforce the need for flexibility in how ethnicity data is interpreted and reported, while also recognising the importance of consistency for statistical outputs. As we continue to refine the harmonised standard, we will consider both the analytical implications and the lived experiences of individuals whose identities may evolve over time.

For example, people sometimes simplify their ethnicity when reporting to reduce respondent burden or complete the task quicker, or their response may be affected by what they feel the interviewer or organisation is looking for (also known as social desirability bias).

Our research highlights that individuals’ willingness to share their ethnic identity and how they choose to describe it, can be influenced by the context in which the question is asked. Factors such as who is collecting the data, how the data will be used, and the perceived trustworthiness of the organisation all play a role in shaping responses.

To support more accurate and confident self-identification, we plan to test optional introductory wording that can be used in data collection settings to build trust and ensure clear purpose. This wording is likely to be presented as a short summary or block of text before the question stem; functioning as a “landing page” to reassure respondents and explain why the data is being collected.

Because trust levels vary across contexts, this wording will remain optional. For example, individuals may feel more comfortable sharing data with government departments than with other organisations. We encourage organisations using the GSS Harmonised Standards to test and tailor introductory text that suits their specific audience and setting.

This approach supports the Inclusive Data Taskforce’s recommendation to improve transparency and build public confidence in how personal data is collected and used.

Examples of proposed changes:

Old question: No information provided to encourage reporting

New question: An optional landing page with information to encourage response

For example, over time and/or geographies or across different ethnicity datasets; our research identifies additional guidance needs, and highlights that comparability remains an important user need when considering changes to the standard.

Improving the inclusivity of the ethnicity harmonised standard will inherently disrupt timeseries data as the questions change. However, we must retain some comparability to the current harmonised standard, as highlighted in User Need 7: output alignment to the census.

Our commitment is to maintain comparability for the top five groups:

• Asian
• Black
• Mixed
• White
• Other

We are working closely with government partners to ensure outputs are appropriate, especially if multiselect options are introduced. Our goal is to deliver a single version or a suite of harmonised, new ethnicity questions for use across the UK. This includes allocating sufficient time and resources to develop questions suitable for different data collection modes, ensuring data coherence.

Examples of proposed changes:

Old question: Different versions across UK nations

New question: Aligned versions developed to better support use across all UK nations

Text presented in a grey box: Stakeholders need more information about various aspects of outputting ethnicity data.

Stakeholders, including statistical producers and data users, have expressed a need for clearer guidance on how to code and aggregate write-in responses under the “Other” category. While this is a valid concern, it is unlikely that further central guidance will be developed at this stage, for the following reasons:

• Limited processing of write-ins: Few data sources currently collect or systematically process write-in responses for the “Other” category. For more detail, see our accompanying work on write-ins for ethnicity.
• Impact of question redesign: The introduction of a new question format will significantly affect how outputs are structured and interpreted, making it premature to provide definitive coding guidance.

The redesigned harmonised standard aims to reduce reliance on the “Other” category by expanding the range and relevance of tick boxes. However, this will require extensive testing at scale to assess the effectiveness of the new design in practice. In the meantime, we encourage organisations to document their own coding approaches where write-ins are used, to support transparency and consistency.

Aggregating granular data meaningfully for analytical purposes remains a key user need; with additional challenges identified for data collectors around processing (coding and aggregating) responses from write-in options; some data collectors and producers are limited in their ability to process or store write-in responses.

While there is a clear need for more granular ethnic group data, implementing this in smaller data collection settings presents practical challenges. As this review of the ethnicity harmonised standard progresses through the Alpha phase, the aggregation of smaller groups will be carefully examined.

Expanding the range of ethnic group options available to respondents aims to reduce reliance on the “Other” categories. New data from Census 2021 shows an increase in respondents selecting “Other” compared to 2011, suggesting that the existing categories may not fully reflect the ways in which people identify. Increased use of the “Other” category can limit the visibility of  smaller or emerging groups, which may affect the inclusivity and utility of the data.

As government efforts intensify to maximise the use of administrative data—particularly from health sources for ethnicity, where write-in responses are often excluded due to cost—it’s essential that the standard’s tick boxes better reflect the UK’s evolving demographic landscape.

Examples of proposed changes:

Old question: Based on England and Wales Census 2021 tick boxes, informed by a prioritisation tool

New question: There will be UK wide consultation to inform updated tick boxes for consistent use across nations

When considering changes to the standard, the time and resource for implementation by data collectors is important (including the lifecycle of survey or administrative data collection).

Implementing changes to harmonised standards, particularly those as sensitive and complex as ethnicity classifications, requires significant lead-in time for data collectors. Our discovery work highlighted that this is often due to the need for system updates, internal approvals, training, and integration of new standards into existing data collection and processing pipelines.

To mitigate these delays, we are engaging early with key government departments to build awareness and readiness for change. This proactive approach is designed to reduce friction later in the process, ensuring that when the final design is agreed, systems and teams are already preparing for adoption. As the question format stabilises, we expect engagement to increase organically across the system.

This phased implementation strategy reflects best practice in change management and aligns with the Government Digital Service’s agile delivery model, which encourages early testing, iteration, and stakeholder involvement to reduce implementation risks. It also supports the Inclusive Data Taskforce’s emphasis on inclusive, coordinated, and well communicated change across the statistical system.

Looking ahead, we will continue to work closely with the Government Statistical Service (GSS) Harmonisation Champions and other partners to ensure the transition to the new ethnicity standard is both timely and effective.

For example, data collectors and processors identified a need to engage with their stakeholders to secure support for implementing any changes; strongly related to user need 4.

The early Alpha phase is intentionally exploratory, designed to “fail fast” and iterate rapidly to uncover and resolve design issues. This approach aligns with Government Digital Service (GDS) guidance, which defines Alpha as a phase for testing multiple prototypes with users and refining solutions based on feedback. As a result, early designs can vary significantly as we refine the question format through successive rounds of testing.

We have engaged with key government stakeholders to validate different options for the question structure to ensure alignment with broader strategic needs. This targeted engagement with influential departments is intended to build early support and facilitate wider adoption across government. We are also collaborating closely with teams across the Office for National Statistics (ONS) to support a smooth transition to the new ethnicity question standard.

We acknowledge that engagement during this phase has been limited in breadth. To date we have prioritised engaging with producers of official statistics. Broader consultation with the wider public sector and community groups has not yet taken place. As we move into the next phase, with a clearer direction for the new standard, we will expand our engagement to ensure the standard is inclusive, robust, and widely supported. This will complement the work on the upcoming public consultation where we will engage with community groups and the public.

Different aspects of ethnicity were important to different users (for example, physical appearance was important for those working on discrimination).

Our Discovery and Alpha research reaffirmed that ethnicity is a deeply personal and self-defined concept. It can reflect a range of factors including cultural heritage, national identity, family history, and for some individuals, physical appearance, particularly in contexts related to discrimination and equalities monitoring.

This complexity presents challenges for both question design and interpretation. Attempts to group similar response options, such as by region or perceived similarity, did not resonate with the public. Many respondents are conditioned by census ethnicity questions and expect to see familiar categories. Moreover, GSS Harmonisation is not intended to define ethnicity itself, but rather support consistency in how data is collected and reported across government.

The subjective nature of ethnicity means that even a single response option can be interpreted in multiple ways. For example, one respondent may select “Pakistani” based on familial or cultural ties, while another may choose it as an expression of national identity. This reinforces the need for clear guidance on what the question is aiming to capture, and how the data will be used.

ONS guidance on measuring ethnic group, national identity, and religion acknowledges this complexity, noting that these are self-identification measures that are “subjectively meaningful to an individual” and may evolve over time. Similarly, the Race Disparity Unit’s standards for ethnicity data emphasise the importance of clarity, consistency, and quality in how ethnicity is recorded and communicated.

As we continue to refine the question design, we will work to ensure that harmonised guidance accompanying the new standard clearly communicates its purpose, limitations, and intended use.

The current standard is widely used and it aligns with the five ‘high-level’ categories in the census question; current ethnicity outputs often align or aggregate to these categories to enable comparison with census or other sources and this will continue to be a user need if any changes are made.

A consistent theme from our engagement with data users is the importance of aligning outputs with the five high-level ethnic group categories used in the Census:

• Asian
• Black
• Mixed
• White
• Other

This alignment is critical for enabling continuity in reporting, benchmarking, and policy evaluation across government. While improvements to the harmonised standard may introduce more inclusive and flexible question formats, particularly through options like multiselect, our commitment remains to preserve output comparability at this top level.

Importantly, our research and user engagement to date have not identified a strong need for additional high-level categories beyond those already established in the Census. This reinforces the value of maintaining a stable output structure, even as the underlying data collection methods evolve.

As we move into later phases of development, we will continue to work closely with the Office for National Statistics (ONS) and other partners to ensure that any new question designs can be mapped reliably to these five categories. This will support coherence across datasets and uphold the integrity of long-term trend analysis.

Some third sector organisations and local authorities need more detailed data to better support groups that are disadvantaged; this must be balanced with the need to aggregate data.

Discovery feedback from local authorities and third sector organisations highlighted a growing demand for more granular ethnicity data. These users often require detailed insights to better understand and serve specific communities, particularly in areas such as service delivery, equality monitoring, and community engagement.

However, increasing granularity introduces important considerations around data disclosure and statistical robustness. Balancing the need for detailed data with the need to protect respondent confidentiality will be a key focus in the next phase of development, particularly as we work through how ethnic groups may be aggregated in the new question.

Some local authorities and community stakeholders have expressed interest in expanding the national identity response options beyond the single tick box of ‘English, Welsh, Scottish, Northern Irish, or British’. National identity is a separate question, which the harmonised standard recommends should always be asked alongside ethnicity and religion to form a cultural identity suite of questions. However, discovery work has shown that this is not consistently implemented across UK surveys and administrative sources.

To address this, we have tested how respondents react to being asked for more detailed national identity information within the ethnicity question, as well as testing this alongside a national identity question. We will continue to explore the overlap between these concepts through further testing and deeper statistical analysis.

Some groups and communities report that identifying within the “Other” category can make respondents feel marginalised or not part of the UK.

Feedback from both data users and the public consistently highlights that the “Other” category can feel exclusionary. Many respondents report selecting it not because it reflects their identity, but because none of the available options feel appropriate. This can lead to frustration, disengagement, and, ultimately, lower data quality.

Reducing reliance on the “Other” category is therefore not just a technical improvement, but a matter of inclusivity and trust. When people do not see themselves represented in the available options, it undermines confidence in the data collection process and the institutions and organisations behind it.

Our redesign work is focused on expanding and refining response options to better reflect the UK’s diverse population. By doing so, we aim to ensure that more people can identify with a listed category, reducing the need to default to “Other.” This is particularly important in administrative data settings, where write-in responses are often not processed, and the “Other” category becomes a catch-all that obscures meaningful variation.

It is not feasible to design a data collection question that captures every possible way individuals may identify, as doing so would create an excessively long list and increase respondent burden. To support inclusivity and flexibility, the new ethnicity question will continue to include an ‘Other’ option.

This user need reinforces the importance of inclusive design and aligns with the Inclusive Data Taskforce’s call to improve the visibility of underrepresented groups in official statistics. It also supports our broader goal of improving data quality, coherence, and public confidence in how ethnicity data is collected and used.