Ethnic group tick-box prioritisation report for Census 2021 in England and Wales

The ethnic group question was first included in a UK Census in 1991. This question has been reviewed and updated each census to reflect the changing profile of England and Wales, and to take account of changes in terminology.

Between June and August 2015, the Office for National Statistics (ONS) conducted a public consultation on topics for Census 2021. This consultation confirmed a need to continue to collect information on ethnic group and highlighted areas where specific needs might not be met by the current question. We then ran an “Ethnic Group Stakeholder Follow-Up Survey” between November 2016 and January 2017 to gain a deeper understanding of user need for ethnic group information. The survey and topic consultation identified requests for 55 new ethnic group tick-boxes to be added to the 2021 Census ethnic group question (Annex B from the ‘2021 Census: Ethnic Group Stakeholder Follow-up Survey – Summary of responses’ includes the full list of requests). We have evaluated these requests through extensive analysis, research, consultation and question testing.

The methodology used to evaluate those requests was published in an earlier report: The ethnic group prioritisation tool: 2021 Census in England and Wales (PDF, 220KB). The ethnic group prioritisation worked by assessing potential tick-boxes against seven principles grouped into five criteria. A description of each of the criteria and principles is provided in Table 2 of that earlier report.

This report shows how that methodology was applied to determine the most appropriate response categories for the ethnic group question in Census 2021. We recognised all requests for ethnic group tick-boxes as valid. The evaluation made no judgement on the legitimacy of groups’ requests. The prioritisation exercise did consider:

• whether there is a need to collect the requested data to support equality monitoring and policy development and/or service delivery or resource allocation,
• whether data of sufficient quality can be collected
• the impact on comparability with previous census data
• whether alternative data are already collected within the census
• whether it is acceptable to explicitly include the group with a tick-box (both to the general public and to the groups it is trying to measure)

The 55 tick-boxes requested were first evaluated for strength of user need using evidence from the consultation and other relevant correspondence and stakeholder engagement. Once the user need evaluation was completed, those requests that had demonstrated adequate user need (that is, Gypsy, Irish Traveller, Sikh, Somali, Jewish, Roma, Korean and Kashmiri) were then evaluated against further criteria (alternative data sources, data quality and comparability). The December 2017 topic report, identified four groups (Sikh, Roma, Jewish and Somali) where more work was needed before we could consider recommending any new additional response categories.

In the 2021 Census White Paper and our December 2018 topic report we recommended the following changes to the Census 2021 ethnic group question:

• include a tick-box for Roma next to the “Gypsy or Irish Traveller” tick-box, within the high-level “White” category
• include a write-in option for those selecting African within the high-level Black African, Caribbean or Black British category to enable a more specific ethnic background to be recorded

The full evaluation that led to these recommendations is outlined in the Evaluation of evidence section.

Annex A details the scoring system for each of the seven principles and includes the evidence used to assess each principle. Annex B summarises the scores and evaluations for the following groups: Gypsy, Irish Traveller, Sikh, Somali, Jewish, Roma, Korean and Kashmiri.

All scoring has been quality assured by the ONS Census 2021 ethnic group, language, national identity and religion Topic Group and the Ethnic Group Assurance Panel (membership of these groups are detailed in Section 3 of Development of the evaluation tool, The ethnic group prioritisation tool: 2021 Census in England and Wales).

Census 2021 will be the most inclusive ever. Write-in answers, online “search-as-you-type” functionality, together with provision of communications, community engagement, and advice and help from field force and contact centre staff, will ensure everyone can tell us how they wish to identify themselves.

To help meet the needs of those groups who do not have a specific tick-box we will:

• continue to provide write-in options for respondents to identify as they wish
• develop “search-as-you-type” functionality, which will make it easier for individuals to identify as they wish
• provide support to communities including community-specific communications, to encourage participation and raise awareness that respondents can identify as they wish
• continue to include a religion question, which will include specific Sikh and Jewish response options (tick-boxes)
• increase the census analytical offering and outputs for religion and ethnic group
• produce analysis for groups including Sikh, Jewish, Gypsy or Irish Traveller and Roma populations to meet the need for high quality data about these groups; we have also made a specific analytical offering for the population who identify as Cornish
• publish univariate outputs (data from a single characteristic) for write-in responses (such as data from the ethnic group write-in responses) for all geographies, where there is a population of sufficient size and an adequate user requirement has been identified
• use the Digital Economy Act 2017 to undertake data-linking for research purposes that will ensure that data on religion is available across public services, not just census-collected data
• update Government Statistical Service (GSS) guidance on harmonisation to ensure that public bodies are fully aware of their duties to record religion data

Additional information on the overall design of the Census 2021 questionnaire can be found in the Question and questionnaire development overview for Census 2021. Details of the extensive research and testing that have supported our decision-making can be found in the Summary of testing for Census 2021.

To evaluate requests for the 55 new ethnic group tick-boxes for Census 2021, we reviewed and updated the prioritisation tool first used in 2011, as set out in The ethnic group prioritisation tool: 2021 Census in England and Wales.

The evaluation process for deciding whether to include new tick-boxes was:

• Step 1: requests for 55 new tick-boxes were evaluated on the “strength of user need” criteria using evidence from the consultation exercise and other relevant correspondence and stakeholder engagement
• Step 2: 8 requests were then evaluated against the next three criteria – “lack of alternative sources of information”, “data quality” and “comparability” using quantitative evidence from testing and from the 2011 Census, including write-in responses; qualitative evidence from the 2015 Topic Consultation, the Ethnic Group Stakeholder Follow-Up Survey and ethnic group literature
• Step 3: 4 requests were then evaluated against the acceptability, clarity and quality criterion, using qualitative evidence from focus groups, workshops and cognitive interviews; quantitative online surveys of acceptability, Gurdwara survey and stakeholder engagement
• Step 4: tick-box recommendation was made after considering all evidence evaluated

Step 1: 55 tick-boxes evaluated for strength of user need

Each of the 55 requests was given a score of 2, 1 or 0 for the two principles within the “strength of user need for information” criterion. The two principles were: strength of need for equality monitoring and policy development; and strength of need for resource allocation and service delivery.

If a request was scored as 0 for both the user need principles, then it was not scored further. Of the 55 requests (listed in the “2021 Census: Ethnic Group Stakeholder Follow-up Survey – Summary of responses” in Annex B), 8 of these requests scored greater than 0 and therefore demonstrated enough user need to require further consideration.

Step 2: 8 requests evaluated against “lack of alternative sources of information”, “data quality” and “comparability” criteria

The eight requests that had a high enough strength of user need to consider further were (alphabetically):

• Gypsy
• Irish Traveller
• Jewish
• Kashmiri
• Korean
• Roma
• Sikh
• Somali

For Gypsy and Irish Traveller, the request was for these groups to be separately identified. In 2011, there was a single tick-box, “Gypsy or Irish Traveller”, that covered both these groups.

These eight requests were then scored 2, 1 or 0 on the principles of “lack of alternative sources of information”, “data quality” and “comparability”. Four of these, which had sufficiently strong case for inclusion, were evaluated further.

Step 3: 4 requests evaluated against the “acceptability, clarity and quality” criterion

The four requests that scored highly enough to evaluate further were:

• Jewish
• Roma
• Sikh
• Somali

In the 2021 Census topic research: December 2017 report, we confirmed we would take forward Jewish, Roma, Sikh and Somali tick-boxes as the top scoring requests to finalise our decisions on the ethnic group categories. We assessed these further by engagement with stakeholders, further research and confirming our recommendations were compliant with any legal obligations.

The requests were assessed using a RAG status (Red, Amber, Green) against the final criterion of acceptability, clarity and quality. Evidence that fed into this included stakeholder engagement and commissioned research.

Step 4: tick-box recommendation

To reach a final recommendation, all evidence from the previous steps was reviewed. While the final decision on whether to include a tick-box was made after the assessment of the acceptability, clarity and quality criterion, consideration was given to this alongside the assessed scores from the other four criteria (user need, lack of alternative sources, data quality and comparability). Our final recommendation was to:

• include a tick-box for Roma, next to the “Gypsy or Irish Traveller” tick-box, within the high-level “White” category
• include a write-in option for those selecting African within the high-level “Black African, Caribbean or Black British” category to enable a more specific ethnic background to be recorded

The evaluation process we have described required evidence gathering from a number of sources, including the conduct of a number of tests. A list of the tests referenced in this report is shown in Annex C and further details are provided in the Summary of testing for Census 2021.

References to tests are provided throughout in the form (Year: Test number). “Year” references the calendar year the test was undertaken, and “test number” is the position of the test within the year considering all testing that took place in that year. For example, the fifth test conducted in 2017 would be referenced as (2017:5).

In this section we provide a summary of the evaluation exercise for the eight groups that were identified in Step 1 as having a high enough strength of user need to warrant further assessment. These groups were:

• Gypsy
• Irish Traveller
• Jewish
• Kashmiri
• Korean
• Roma
• Sikh
• Somali

For each group, a summary of the evidence that supported the evaluation is given, covering all four of the steps that were outlined in Evaluation process. A table showing the evaluation scores for each of these is given in Annex B.

Separate tick-boxes for Gypsy and Irish Traveller

Background

A joint “Gypsy or Irish Traveller” tick-box was added to the ethnic group question under the high-level “White” category in the 2011 Census ethnic group question. Stakeholders have requested that separate information on Gypsies and Irish Travellers be captured on Census 2021.

57,680 individuals in the 2011 Census in England and Wales were recorded as Gypsy or Irish Traveller within the White high-level category (including those who used the tick-box and those who used the White write-in option).

Criterion: Strength of user need (Step 1)

The 2021 Topic Consultation and the Ethnic group follow-up survey provided some evidence that separate tick-boxes for Gypsies and Irish Travellers are of particular policy interest, with one respondent to the ONS stakeholder engagement exercises stating:

“There is particular interest in the housing circumstances of Gypsy and Irish Travellers…and also in other circumstances [ ] (education, health social disadvantage) in order to develop policies…a further breakdown of ethnicities [of the ‘Gypsy or Irish Traveller’ tick-box] may lead to a more representative figure” Welsh Government.

However, there is little evidence that separate information on Gypsy and Irish Travellers is needed for service delivery or resource allocation.

Criterion: Lack of alternative sources of information (Step 2)

In the 2011 Census, 3% of the Gypsy or Irish Traveller respondents wrote in their separate “Gypsy” or “Irish Traveller” identity within the write-in option, the majority ticking the combined tick-box. This suggests that the majority of this population will identify with the combined tick-box and the Gypsy and Irish Traveller populations are unlikely to use the write-in option.

It is hard to get enough information about the separate “Gypsy” or “Irish Traveller” population, where there is a combined “Gypsy or Irish traveller” tick-box as most people from these groups will use the tick-box, rather than the write-in option.

Criterion: Data quality of information collected (Step 2)

The lack of separate tick-boxes for “Gypsy” and “Irish Travellers” is unlikely to affect the quality of data collected on the Gypsy or the Irish Traveller population. Without a separate “Gypsy” or “Irish Traveller” tick-box, respondents are unlikely to be confused or burdened, as there is an obvious alternative response. From the 2011 Census, the majority of this population are likely to respond consistently by ticking the combined “Gypsy or Irish Traveller” tick-box.

Criterion: Comparability with 2011 data (Step 2)

The inclusion of separate “Gypsy” and “Irish Travellers” tick-boxes would not affect comparisons over time. If the 2011 Census “Gypsy or Irish Traveller” tick-box was split for 2021, the data could still be aggregated (by combining outputs for the separate “Gypsy” and “Irish Traveller” tick-boxes) to produce outputs comparable with 2011 Census data.

Separate Gypsy and Irish Traveller tick-box overall scoring exercise and evaluation (Steps 2 to 4)

The separate Gypsy and Irish Traveller tick-boxes both had a weighted score of 8, as shown in the summary of scores from prioritisation exercise. This was not considered a strong enough case to consider further testing to evaluate acceptability, clarity and quality (Step 3). We recommend that the combined “Gypsy and Irish Traveller” tick-box is retained for the Census 2021.

Jewish

Background

A religious affiliation question, which included a Jewish tick-box, was added to the 2001 Census. This question was included on the 2011 Census and has been recommended for inclusion in Census 2021.

In the 2011 Census for England and Wales, 263,346 people declared their religion as Jewish and 33,770 wrote in Jewish as their ethnic group.

Criterion: Strength of user need (Step 1)

The ONS found little evidence to support the need for a separate Jewish ethnic group tick-box, in addition to the existing Jewish religion tick-box, for equality monitoring or policy development.

There is some evidence that the Jewish population are a group of particular interest for service delivery. Evidence from the stakeholder engagement highlights how data for this group from an ethnic group tick-box would be used nationally and locally to optimise service provision, including for synagogues, Jewish schools, nurseries and care providers.

Criterion: Lack of alternative sources (Step 2)

Table 1 shows estimates of the Jewish population from 2011 Census data for England and Wales, as measured by the religious affiliation or ethnic group questions.

Ethnic group write-in answers are likely to be inadequate for measuring the “Jewish” population as a whole. In 2011, 33,770 people in England and Wales wrote in “Jewish” in response to the ethnic group question, while 263,346 people were recorded as “Jewish” as their religion.

In the 2011 Census, most people with a Jewish religious affiliation, identified in the ethnic group question as “English/Welsh/Scottish/Northern Irish/British” within the White high-level category. Jewish ethnic group write-ins are inconsistent in how they are recorded at the high-level ethnic group categories (Jewish is not consistently written under the “White” category, but spread across the high-level categories). This suggests that any one write-in option of “Jewish” (within White, Black, Asian or Other high-level category) would be a poor measure of this group.

Those who listed their ethnicity as Jewish, but did not list Jewish as their religious affiliation, represented only 3% of the total combined ethnic and religious Jewish population.

The ONS concludes that the religious affiliation question already captures data on the Jewish population and offers a reasonable alternative data source. Those people who may wish to identify only as ethnically Jewish, and not religiously Jewish, are still able to use the ethnic group question write-in option to do so.

Criterion: Data quality of information (Step 2)

Without a dedicated “Jewish” ethnic group tick-box, the quality of information is moderate, as respondents who wrote in “Jewish” within the ethnic group question have done so under more than one high-level ethnic group category. In 2011, the majority of those who identified their ethnic group as “Jewish” used the write-in box within the high-level “White” category. However, 32% of those who self-identified their ethnic group as “Jewish”, did so in the ethnic group “Other” write-in box.

Criterion: Comparability with 2011 Census data (Step 2)

Inclusion of a dedicated “Jewish” tick-box could affect comparisons over time. In the 2011 Census, of those who answered “Jewish” to the ethnic group question, 64% wrote this in the high-level “White” write-in category, and 32% wrote this in the high-level “Other” write-in category. Including a Jewish tick-box in one high-level location (“Other” or “White”) could have an impact on comparisons over time, affecting:

• comparability of the “White” and “Other ethnic group” high-level data

• comparability of the “Any other White” and “Any other ethnic group” write-in estimates

Jewish tick-box overall scoring exercise (Step 2)

The Jewish tick-box had a weighted score of 9.5. We announced in the December 2017 update that we needed to undertake further work before we could decide whether to recommend any new response options, including a Jewish tick-box.

Criterion: Acceptability, clarity and quality (Step 3)

Focus groups (2018:13) with Jewish respondents found a strong consensus that a specific Jewish ethnic group response option is unacceptable. Community leaders also raised concerns about asking both a religion and ethnic group question, and the impact it might have on comparability with 2011 Census data.

Jewish tick-box overall evaluation (Step 4)

We do not recommend the inclusion of a Jewish tick-box in the ethnic group question.

Kashmiri

Background

Ahead of the 2011 Census, the ONS conducted a Kashmiri Research Project (October 2009) to consider the need for a Kashmiri tick-box in the 2011 Census ethnic group question. This research showed that differences in deprivation as measured by the census socio-economic questions, between Kashmiri and Pakistani communities were small. Research also showed Kashmiris did not have strong feelings about the need for a tick-box. There were also concerns among respondents over which box to tick if there was also a Kashmiri tick-box.

In the 2011 Census, 25,335 respondents identified as Kashmiri using the ethnic group question write-ins.

Criterion: Strength of user need (Step 1)

There is some evidence from correspondence that Kashmiri is a group of interest for equality monitoring and policy development. Correspondence highlighted the low take-up of services from this group and the need to understand this population. However, the ONS has received little evidence of the need for a Kashmiri tick-box in the ethnic group question for resource allocation or service delivery.

Criterion: Lack of alternative sources of information (Step 2)

The high-level ethnic group “Asian” write-in option could provide an estimate for the Kashmiri population. However, some stakeholders have suggested that this would undercount the Kashmiri population. Without a dedicated “Kashmiri” tick-box some of this group write in, while others may tick, the “Pakistani” tick-box.

The 2011 Census suggests that those that write in are likely to do so in a single location: “Any other Asian background”. As a result, write-in answers could be used as an alternative source of data. There are no other census questions that can act as a proxy for the Kashmiri population.

Criterion: Data quality of information collected (Step 2)

There would be little improvement to the quality of the high-level ethnic group data if a dedicated Kashmiri tick-box was added. 98% of those who wrote in Kashmiri, did so in the high-level “Asian/Asian British” category, “Any other Asian background, write-in” box, with only 1% writing in the “Other” ethnic group category.

If the tick-box was added, it would not be mutually exclusive, as respondents may identify with more than one tick-box (such as Kashmiri and Pakistani or Indian). This could cause more confusion.

Criterion: Comparability with 2011 Census data (Step 2)

The inclusion of a Kashmiri tick-box would not affect comparisons of the higher-level categories with previous census data as this group mainly use the “Any other Asian background” option to record themselves.

Kashmiri overall evaluation (Steps 2 to 4)

The Kashmiri weighted score is 9.5 overall. In 2009, we conducted a large-scale evaluation that provided additional evidence to consider. The Kashmiri Research Project showed that participants were sometimes unsure of how to respond when a Kashmiri tick-box was present. Considering this research, and Census 2021 ethnic group prioritisation score together, there was not a strong enough case to justify conducting further focus groups to consider the acceptability, clarity and quality of adding an additional tick-box.

There was not a strong enough case to recommend a Kashmiri tick-box in Census 2021. Many arguments for data related to Kashmiri are linked to the need to ensure services are provided in appropriate languages and can therefore be better dealt with by the language question, which was added in 2011.

Korean

Background

In the 2011 Census, there were 16,586 South Korean-born residents in England and Wales, compared with 381 North Korean-born residents.

22,514 respondents identified as Korean using the ethnic group question write-ins.

Criterion: Strength of user need (Step 1)

There is some evidence that Korean communities experience disadvantage in health. A respondent to the ONS stakeholder engagement exercises stated:

“We are currently undertaking a Joint Strategic Needs Assessment on the Korean community, which has been identified as a community facing poorer health outcomes, particularly around mental health.” Royal Borough of Kingston upon Thames

There is also some evidence that Koreans are of particular interest for service delivery and resource allocation:

“The Council has … specifically required data on the size of the Korean […] populations in Kingston for a number of purposes. … Data on the Korean population in Kingston was … used for a bid that secured a £240,000 European Integration Fund grant from the Home Office, partly to employ a Korean Link Worker. Working with Kingston College and Learn English At Home, the aim of the project was to integrate and improve the health and well-being of people who are often isolated and vulnerable within the community, and to support people by breaking down language barriers. In this instance, the target population was estimated using language data from the 2011 Census. However, again the Council would have benefited from full census data on the Korean ethnic group.” Royal Borough of Kingston upon Thames

Criterion: Lack of alternative sources of information (Step 2)

Write-in answers could offer a good estimate of the Korean population. Without a dedicated Korean tick-box, most of this group are likely to write in, and to do so with the same response in the same location. In 2011, the majority of individuals who wrote in “Korean” (97%) did so using the “Any other Asian background” write-in option. This group is unlikely to identify with any of the existing “Asian or Asian British” tick-boxes (for example, “Indian”, “Pakistani”, “Bangladeshi” or “Chinese”).

In 2011 Census data, 66% of those who wrote in “Korean” within the ethnic group write-in options, also wrote “Korean” in the national identity question. This other census information could be used as a source of information about the Korean population within England and Wales. The national identity question was the best proxy available. Further analysis showed that data on the Korean population was also collected in the country of birth and passports held questions.

Criterion: Data quality of information collected (Step 2)

Without this tick-box, respondents are unlikely to be confused or burdened, as there is an obvious single alternative for this group (the “Any other Asian background, write-in” option). As there is a clear alternative, data quality is likely to be good. In addition, adding a Korean tick-box is unlikely to improve data quality as the majority of Korean people (97%) already respond within the same high-level category.

Criterion: Comparability with 2011 Census data (Step 2)

We believe that the inclusion of a Korean tick-box will not affect comparisons over time, as this group mainly used the “Any other Asian background” option to record themselves. All outputs (except the “Any other Asian background” count) would be comparable with the 2011 data if a Korean tick-box was included, and any change in the “Any other Asian background” count can be reliably estimated.

Korean tick-box overall scoring exercise and evaluation (Steps 2 to 4)

The Korean tick-box had a weighted score of 8.5. This was not considered a strong enough case to consider further testing to evaluate acceptability, clarity and quality. We do not recommend adding a Korean tick-box to Census 2021.

Roma

Background

In the 2011 Census, 730 people recorded themselves as Roma through the ethnic group write-in options. The University of Salford estimates that the Roma population living in England and Wales was 194,000 in 2012.

Evidence from stakeholders also highlighted different user need for population data on the Roma population when compared with data on the Gypsy or Irish Traveller population. The combined data would be less useful, making it harder for service providers to identify their target population. Because of this, we have only considered the case for a separate “Roma” tick-box, and not for a combined “Roma, Gypsy and Irish Traveller” tick-box.

Criterion: Strength of user need (Step 1)

There is strong evidence that Roma communities are experiencing significant disadvantage across several areas of life. Evidence included the need to develop policy in relation to considerably poorer outcomes of Roma children with life threatening illnesses being presented very late.

There is also evidence that Roma are of interest for service delivery and resource allocation. Respondents to the stakeholder engagement exercises stated:

“…ability to identify Roma regardless of nation of origin would also be extremely useful ... allowing us to produce better projections and school place planning … to understand language needs and design services appropriately.” Oldham Council

Criterion: Lack of alternative sources of information (Step 2)

Write-in answers on the ethnic group question are likely to be inadequate for recording the number of Roma in England and Wales. Without a “Roma” tick-box, very few Roma are likely to write in. The Roma population was estimated to be 194,000 in England and Wales in 2012; however, in the 2011 Census only 730 people wrote in Roma as their ethnic group.

No other census question could be used as a proxy for the Roma population. Roma migrate from a variety of Central and Eastern European countries, so country of birth and passports held would be poor indicators of this population. Roma populations do not speak a common language distinct from other ethnic groups, so the language question would not be a good proxy.

Criterion: Data quality of information collected (Step 2)

Without a dedicated Roma tick-box, the Roma population may be unsure of where to answer, reducing the quality of information collected. In the 2011 Census, those who wrote in Roma did so in a number of places (for example, under “Any other White background” and “Any other ethnic group”). There is no obvious single write-in location.

Without a tick-box, Roma may also identify as Gypsy and tick the “Gypsy or Irish Traveller”, or may be more likely to identify in other ways such as Slovakian or Hungarian.

Lower literacy within parts of the Roma population has been documented. The addition of a tick-box could improve data about the Roma population by simplifying the process of answering the census for this group.

Criterion: Comparability with the 2011 Census data (Step 2)

The inclusion of a “Roma” tick-box may affect comparisons over time to some extent. The Roma write-in answers from the 2011 Census were small, only 730 Roma, which is considered a severe undercount of the population. This indicates that this population are either not answering the census or not recording their Roma identity, and may have identified in an alternative way, such as a Hungarian or Slovakian instead.

There may also be a small impact on comparability over time if people who previously responded with identities such as Romanian or Slovakian, change to ticking a Roma tick-box. This could cause some movement in data; however, this is likely to be small.

Roma tick-box overall scoring exercise (Step 2)

Roma had a weighted score of 16.5. This was considered a strong enough case to consider further testing to evaluate acceptability, clarity and quality.

Criterion: Acceptability, clarity and quality (Step 3)

Focus groups (2018:13), workshops (2018:5) and interviews conducted with the Roma population found the term Roma was acceptable and the majority wanted a tick-box to identify themselves. Many Roma respondents did not identify with the term “Gypsy”. The addition of a “Roma” tick-box eased the burden on respondents. Without the tick-box respondents were uncertain about which high-level category they should select.

The research (2018:25) suggested that the most acceptable location of the tick-box was within the higher-level “White” category. Placing it there, after the “Gypsy or Irish Traveller” tick-box, made it easy for the Roma respondents to find. We considered how acceptable the term “Roma” was to the general population (2018:16). 86% of those who responded were comfortable with the term “Roma” being included within the ethnic group question.

Roma tick-box overall evaluation (Step 4)

We recommend including a Roma tick-box.

Sikh

Background

A religious affiliation question was added to census from 2001. One of the aims of this question was to collect data on religious minorities that cut across national and ethnic boundaries. A “Sikh” tick-box was included for this question in the 2001 and 2011 censuses and will be included in the Census 2021 questionnaire.

In the 2011 Census in England and Wales, 423,158 people identified their religious affiliation as Sikh. 83,362 wrote in “Sikh” in the ethnic group question, of which the majority (76,500) also ticked Sikh within the religion question.

Criterion: Strength of user need (Step 1)

There is evidence that Sikhs are experiencing significant disadvantage in multiple areas of life, including employment, pay and educational attainment. For example, respondents to the ONS stakeholder engagement exercises stated that: “Sikhs are more likely to be unemployed, less well represented in top status jobs, and less well paid”’ Sikh Federation UK and Sikh Network

There is also some evidence that Sikhs are of particular interest for resource allocation and service delivery, with stakeholders stating that:

“What we would benefit from would be the inclusion of a Sikh category within the current Ethnic categories ... Governors, teachers and educational professionals such as ourselves often use ethnicity data to review emerging and well-established trends. This gives us rich information to target interventions, development support and seek funding arrangements from Local Educational Authorities and Government.” University of Wolverhampton.

Criterion: Lack of alternative sources (Step 2)

Table 2 shows 2011 Census data for England and Wales, of those who recorded their religious affiliation and/or ethnic group as Sikh.

In the 2011 Census, 83,362 people wrote in “Sikh” in the England and Wales ethnic group question, while 423,158 people ticked “Sikh” in the religious affiliation question.

74% of people who ticked “Sikh” in the religion question ticked “Indian” on the ethnic group question.

The religion question offers a good proxy for the ethnic Sikh population. In the 2011 Census, a majority of the population who identified as Sikh in the ethnic group question (92%) were captured by the religious affiliation question.

The 2017 ethnic group question test (2017:3) supported analysis of the Census 2011 data, finding that there is no indication that the religious affiliation and ethnic group questions are capturing different Sikh populations. Those people who may wish to identify only as ethnically Sikh, and not religiously, are still able to use the ethnic group question write-in options to identify as they wish.

Criterion: Data quality of information (Step 2)

Without a Sikh tick-box in the ethnic group question, people wishing to identify as Sikh may be unsure of where to answer, reducing data quality. Analysis of the 2011 Census ethnic group question found that those who wrote in Sikh did so in different places: 51% under the high-level “Any other ethnic group” and 48% under “Any other Asian background”. This resulted in a total combined estimate 83,362 people who wrote in “Sikh” within the ethnic group question.

However, further research detailed in the criteria in this section found the tick-box was not acceptable to some respondents and suggested that the majority of those answering Sikh to the religion question ticked the “Indian” ethnic group tick-box, even when a Sikh tick-box was added

Criterion: Comparability with 2011 Census data (Step 2)

Analysis of 2011 Census data and the 2017 ethnic group question test (2017:3) found that the inclusion of a Sikh tick-box in one location (“Other” or “Asian/Asian British”) could have an impact on comparisons over time affecting:

• comparability of the “Asian/Asian British” and “Other ethnic group” high-level data
• comparability of the “Any other Asian background” and “Any other ethnic group” write-in estimates
• comparability of the “Indian” tick-box estimate

Sikh tick-box overall scoring exercise (Step 2)

The Sikh tick-box scored a weighted score of 12. This was considered a strong enough case to consider further testing to evaluate acceptability, clarity and quality (Step 3).

Criterion: Acceptability, clarity and quality (Step 3)

The majority of those answering Sikh to the religion question in the 2011 Census and the 2017 ethnic group question test ticked the “Indian” ethnic group tick-box. This was still the case when a Sikh tick-box was added to the ethnic group question in testing.

Focus groups (2018:13) commissioned by the ONS considered the inclusion of a Sikh tick-box, explored how Sikh respondents answered the ethnic group question and explored the acceptability, clarity and impact on the quality of the data when a Sikh tick-box was added.

This research concluded that:

• the addition of a Sikh tick-box to the ethnic group question was seen as unacceptable to some participants, particularly among younger, second-generation participants

• the inclusion of a Sikh tick-box under “Asian” caused confusion over having to choose between an “Indian” or “Sikh” identity

• including “Sikh” under “Other ethnic groups” prompted concerns and suspicions about why Sikh was being separated from the Asian or Indian tick-boxes

• there was a feeling that if one religion is included under ethnic group, all religions should be

The ONS received information from a survey of Gurdwaras enquiring about acceptance of a Sikh ethnic group tick-box, which showed a high acceptance for inclusion. The survey gave the ONS more insight into the views of Sikh groups, alongside the ONS’s other research. Independent research was undertaken for the ONS to further understand the acceptability of the Sikh response option within the ethnic group question. The research showed that there are differing views within the Sikh population about whether a specific tick-box should be added to Census 2021 ethnic group question.

Sikh tick-box overall evaluation (Step 4)

There is not a strong enough case to include a Sikh tick-box. Quantitative analysis showed that there is no evidence that the religious affiliation and ethnic group questions are capturing different Sikh populations. Qualitative research found that the addition of a Sikh tick-box caused confusion and was unacceptable to some participants, particularly among the second generation.

We will continue to include a Sikh tick-box for the religion question in the census and, more widely, the UK Statistics Authority will strengthen the harmonisation guidance on the collection of data on religion alongside ethnicity data across government.

Somali

Background

In the 2011 Census in England and Wales, 45,475 wrote in their ethnic group as “Somali”.

Criterion: Strength of user need (Step 1)

There is strong evidence that there is a need to understand data from ethnic “Somali” populations. Evidence suggested this population is experiencing significant disadvantage in several areas of life, including employment, housing, health and education. There is also evidence that Somalis are of particular policy interest. Respondents to the ONS stakeholder engagement exercises stated that:

“In Tower Hamlets, we require information specifically about the Somali population… we have used proxy data, using country of birth data from the 2011 Census, to profile the first-generation Somali population. This analysis has quantified the severe disadvantage this population faces, in terms of poorer outcomes across employment, housing, health and education. For example, only one third of the working age Somali-born population were in employment in 2011, compared with almost two thirds of the borough population (34 vs. 64 %)…Having Census data on the (ethnically) Somali population, would allow the Council to… undertake policy evaluation: i.e. evaluate the success of policies and actions, which aim to target the poor outcomes the Somali population faces.” London Borough of Tower Hamlets

There is also some evidence that the Somali population is of particular interest for service planning and delivery, with stakeholders stating:

“The statistics help to justify equalities monitoring when we record who uses services provided by the council or commissioned providers [ ] to see if there are different levels of service uptake or differential outcomes e.g. health, sexual assault and hate crime. Ethnic group statistics are also used as a predictor of future needs (particularly health needs). There are currently no alternative sources for this information.” Bristol City Council

Criterion: Lack of alternative sources (Step 2)

Without a “Somali” tick-box, some Somali residents may write in “Somali” under the ethnic group question, while others may select the “African” or “Arab” tick-boxes. Those that write in are likely to write the same response in the same location (in 2011 most used the “Any other Black/African/Caribbean background” write-in option). Write-in answers could be used as an alternative data source about Somali communities. However, as some who could identify as ethnically Somali may also select other identities, the estimate may not fully meet the needs of data users.

No other census question could be used as an accurate proxy for the Somali population. Based on the 45,475 respondents in the 2011 Census who identified as Somali in the ethnic group write-in responses, only 53% were captured by the main language question, which offers the best proxy, while country of birth captured 49%. The fact that the Somali population is entering second and third generations causes proxy data on main language and country of birth to be even less useful as an estimate of the size of the ethnically Somali population.

Criterion: Data quality of information (Step 2)

Without a “Somali” ethnic group tick-box, some respondents, who wish to identify as Somali, may be unsure of where to answer, reducing the quality of information collected. In the 2011 Census, 83% of Somalis who wrote in did so under “Any other Black/African/Caribbean background” and 14% wrote in under “Any other ethnic group” within the high-level “Other ethnic group” category. As the majority of those that identified as Somali (that is, those who wrote in) wrote in the same place, this can provide some useful data.

However, testing shows that many Somalis do not write in a specific African identity and may tick the “African” tick-box (2018:13). This could reduce the granularity of the data and not provide data users with the level of detail they have requested.

Criteria: Comparability with 2011 Census data (Step 2)

Inclusion of a dedicated “Somali” tick-box could affect comparisons over time; however, the size of this shift is likely to be small in the higher-level categories. In 2011, most (83%) Somalis who wrote in did so under “Any other Black/African/Caribbean background”. Some Somalis (14%) wrote in under “Any other ethnic group”, within the “Other ethnic group” high-level category. The addition of a “Somali” tick-box could therefore cause people to shift from the “Other ethnic group” high-level category to the new Somali tick-box, most likely located in the “Black, Black British, Caribbean or African” high-level category.

Somali tick-box overall scoring exercise (Step 2)

The Somali tick-box scored a weighted score of 13.5. This was considered a strong enough case to consider further testing to evaluate acceptability, clarity and quality (Step 3).

Criteria: Acceptability, clarity and quality (Step 3)

Focus groups, commissioned by the ONS, found that the inclusion of a Somali tick-box, without other African ethnicity tick-boxes, was viewed as unacceptable by some participants. The singling out of Somali for a tick-box raised suspicions about why this was being asked and resulted in participants feeling they had to choose between being African or Somali. Within further engagement with community groups and stakeholders there was a strong consensus that better data for the ethnically Somali population is needed. Stakeholders also raised similar concerns that the inclusion of a Somali tick-box, without other African tick-boxes, could raise suspicions and lead to lack of clarity with how to answer.

As a tick-box was viewed as unacceptable by some research participants, the user need for data is high and there is no good proxy data available, we reviewed the “Black, African, Caribbean or Black British” higher-level category to consider the possibility of collecting more information on the Somali population alongside other African identities.

We conducted cognitive testing (2018:30) of new question versions among respondents who identify with “Black British”, “African”, “Caribbean” or “Somali”. This provided evidence on the best way to obtain more detailed data on ethnically Somali individuals. Census 2021 will provide an option for those selecting “African” to allow people to write in specific African ethnicities within the high-level “Black, Black British, Caribbean or African” category.

Additional evidence was provided through our Ethnic Group Assurance Panel, which added more weight to the need for more detailed information for the “Black African” ethnic group. This solution to the collection of “Somali” data will also meet this need to provide a further breakdown of African ethnicities.

Somali tick-box overall evaluation (Step 4)

We do not recommend the inclusion of a separate “Somali” tick-box. We recommend the inclusion of a write-in option for those selecting “African” within the high-level “Black, African, Caribbean or Black British” category. This will allow all those who select “African” to specify their ethnic background further.

As a result of this evaluation, the Office for National Statistics (ONS) recommends that Census 2021 should include:

• a tick-box response option for “Roma” in the high-level “White” category in the ethnic group question
• a write-in option for those selecting “African” to allow people to specify African ethnicities in the high-level “Black, Black British, Caribbean or African” category

These recommendations are set out in the 2021 White Paper and December 2018 topic report.

For other groups considered in this report, the ONS will provide a write-in option online and on paper. A new “search-as-you-type” facility for write-in answers online will include many ethnic group categories including Gypsy, Irish Traveller, Somali, Sikh, Kashmiri, Korean and Jewish. Respondents will continue to be able to state identities that are not listed, as the write-in option online will also accept free-text responses. The paper form write-in option will continue to enable people to identify as they wish in the ethnic group question.

When required, respondents will also be supported in completing the questionnaire in a variety of languages, including, for example, Polish, Punjabi and Somali (through guidance booklets and telephone interpretation service). Census engagement activities aiming to promote completion of the questionnaire, and to encourage respondents to identify as they wish in the ethnic group question, will also have a focus on people from a wide range of different ethnic groups. For example, we will assist Roma organisations to provide support for local communities and to raise awareness of the Roma response option.

The five criteria used to evaluate ethnic group tick-boxes were divided into seven principles. For each of these we describe:

• what the principle is and the rationale for including it
• what evidence was included in the assessment
• the scoring system used

Criterion 1: Strength of user need for information on ethnic group

Principle 1.1

Group is of particular interest for equality monitoring and/or for policy development (for example, group is particularly vulnerable to disadvantage).

Rationale: The census needs to provide ethnic group data for policy development. This enables government departments and local authorities to fulfil their requirements under the Equality Act 2010.

Evidence included: Qualitative evidence from the 2015 Topic Consultation, Ethnic Group Stakeholder Follow-Up Survey and stakeholder engagement.

High score (2): Strong evidence or indication that the group is experiencing significant disadvantage in one or more areas of life and/or that there is particular policy interest.

Medium score (1): Some evidence or indication that the group is experiencing some disadvantage and/or that there is particular policy interest.

Low score (0): Little evidence that this group experiences disadvantage and/or that there is particular policy interest.

Principle 1.2

Group is of particular interest for service delivery and/or resource allocation.

Rationale: The census needs to provide ethnic group data to allow services to be tailored and resources to be allocated. This enables government departments and local authorities to fulfil their requirements under the Equality Act 2010 and to allow services to be tailored and resources to be allocated.

Evidence included: Qualitative evidence from the 2015 Topic Consultation, the Ethnic Group Stakeholder Follow-Up Survey and stakeholder engagement.

High score (2): Strong evidence that the group is of particular interest for service delivery and/or resource allocation.

Medium score (1): Some evidence or indication that the group is of particular interest for service delivery and/or resource allocation.

Low score (0): Little evidence that this group is of particular interest for service delivery and/or resource allocation.

Criterion 2: Lack of alternative sources of information

Principle 2.1

Write-in answers are not adequate for measuring this group.

Rationale: If the majority of the group wrote in answers in a consistent manner, these data could be analysed to provide data without the need for a tick-box.

Evidence includes: Quantitative evidence from the 2011 Census and testing. Qualitative evidence from the 2015 Topic Consultation, the Ethnic Group Stakeholder Follow-Up Survey and ethnic group literature.

High score (2): Without a tick-box very few people are likely to write in and/or they are unlikely to write in the same place and/or they are unlikely to write in consistently; to a degree that write-in response could not be used as a proxy.

Medium score (1): Without a tick-box, low response rates, and inconsistency in response locations and content, mean that write-in responses could be used as a proxy but with some margin of error.

Low score (0): Without a tick-box, the majority of this group are likely to write the same response in the same location.

Principle 2.2

Other census information is inadequate as a suitable proxy (for example, country of birth, religion, national identity, citizenship, and main language).

Rationale: If one or more other census questions provided similar information there is a lesser need to include a tick-box in the ethnic group question.

Evidence includes: Quantitative evidence from the 2011 Census.

High score (2): No other census question could be used as a proxy.

Medium score (1): A large proportion of this group could be captured in another question offering proxy data.

Low score (0): The group will be captured almost entirely by a single alternative census question.

Criterion 3: Data quality of information collected

Principle 3.1

Without this tick-box, respondents would be unduly confused or burdened and so the quality of information would be reduced (for example, if a large, well known, or highly distinct group was left out, and respondents from this group ticked a variety of options instead).

Rationale: If some tick-boxes are expected but left out, respondents that would have ticked one may end up ticking inconsistently as there may not be an obvious available option. Consistent response is required to maximise data quality.

Evidence includes: Analysis of write-ins from 2011 Census.

High score (2): There is no obvious single alternative for this group. Evidence suggests that the group wrote in a number of places. Ticking alternative options would reduce the quality of the data.

Medium score (1): The majority of this group select a single alternative.

Low score (0): Obvious single alternative option (including “other” options) for this group.

Criterion 4: Comparability with 2011 data

Principle 4.1

There will be no adverse impact on comparability.

Rationale: The ONS 2015 Topic Consultation revealed a need for comparability with 2011 data, to enable users to see changes over time.

Evidence includes: Quantitative evidence from the 2011 Census and question testing, qualitative evidence from the 2015 Topic Consultation and the Ethnic Group Stakeholder Follow-Up Survey.

High score (2): Inclusion will not affect comparisons over time, for example, when this population mainly used the “Other” tick-boxes to describe themselves previously.

Medium score (1); Inclusion will affect comparisons over time to some extent, but the effect on comparability can be reliably estimated.

Low score (0): Inclusion will have major effects on comparability over time.

Criterion 5: Acceptability, clarity and quality

Principle 5.1

The addition of the tick-box and/or revised terminology is acceptable to respondents, clear (both in wording and in the context of the question, for example, mutually exclusive categories), and provides the required information to an acceptable level of quality.

Rationale: Inclusion of the tick-box will elicit a high and consistent response and will provide a dataset that reasonably represents a distinct population.

Evidence includes: Qualitative evidence from focus groups, workshops and cognitive interviews. Quantitative online surveys of acceptability, Gurdwara survey and stakeholder engagement.

Green: Evidence that the tick-box is clear and acceptable to respondents. The target for the tick-box is collecting data for all those it is intended for. There is agreement between community leaders that a tick-box should be included.

Amber: Evidence that tick-box is not clear or acceptable to some respondents and that there is some confusion. There is no strong opinion either way, or there are disagreements regarding the inclusion of the tick-box by community leaders.

Red: Evidence that tick-box is not clear or acceptable to respondents. Respondents may express confusion, feel uncomfortable with the terms and many of whom the tick-box is intended may prefer to tick an alternative. Community leaders agree the tick-box should not be included.

A summary of all the scores used in this prioritisation report is available in the Related downloads of this release.

References to tests take the form (Year: Test number). “Year” refers to the calendar year the test was undertaken in and the test number is the position of the test within the year considering all testing that took place in that year. For example, the fifth test conducted in 2017 would be referenced as (2017:5).

A full description of each of these items can be found in summary of testing for Census 2021.