Our recommendations: how can we be more inclusive in our data?
1. Create an environment of trust and trustworthiness which allows and encourages everyone to count and be counted in UK data and evidence.
- 1.1 Trust is crucially important for the collection and use of data and for inclusion in statistics. People are happy to provide their personal information when they believe that (1) their data matters and will be used to improve people’s lives and are convinced of the (2) reliability, (3) responsiveness, (4) openness and inclusiveness, (5) integrity and (6) fairness of the data producer. To enhance trust and trustworthiness in the provision and use of data, data producers should develop a social contract with those who provide their data (the respondents). This should include:
- 1.1.1 a clear explanation of why the data are being collected and how they will be used
- 1.1.2 the confidentiality and anonymity of respondents, if and why their information will be shared with third parties and under what circumstances, if any, de-anonymisation might occur
- 1.1.3 the provision of timely, free and accessible feedback to respondents
- 1.1.4 engaging with relevant groups and populations across the whole data process, seeking their advice and support with conceptualisation and planning, data collection, analysis and distribution
- 1.1.5 the public interest should prevail over organisational, political or personal interests at all stages in the production, management and dissemination of official statistics
This will help to address the most important issues for participants of data collection, to ensure that there are demonstrable benefits, and that the risks and costs to participants have been minimised.
- 1.2 Data producers should work together to undertake long-term engagement activities with relevant groups and populations in order to maintain open dialogue and build trustworthiness. This could be achieved through outreach, local-level knowledge building and recognition, reporting costs and benefits of engaging with data collection activities, and learning from previous data collection activities to address the costs and barriers to participation, such as the 2021/22 Censuses (see also recommendation 1.4)
- 1.3 Data producers should facilitate trust among potential participants and demonstrate their own trustworthiness by increasing diversity among their staff, including those directly collecting data from the public, and by ensuring that participants are all treated with equal respect.
- 1.4 Data producers should undertake appropriate research to identify the practical barriers to participation and implement best practice in data collection, including ethical considerations, to enhance the inclusiveness of the approaches taken. This might entail providing internet access to address the barriers for digitally excluded groups, and translators for those not fluent in English.
- 1.5 Data producers should ensure that data collection instruments are accessible to all, recognising differences in language, literacy, and the relative accessibility of different modes and formats. For example, using multi-mode surveys as standard practice and implementing additional adjustments to enable the participation of adults and children with a range of disabilities, and those who experience other forms of exclusion, including digital exclusion.
- 1.6 Data producers should avoid the use of proxy responses and ensure that the default approach is for self-reporting of personal characteristics, including, where appropriate, collecting information directly from children.
- 1.7 Practical barriers to the access and use of ensuing data should be investigated, as well as ways of promoting confidence in these data.
2. Take a whole system approach, working in partnership with others to improve the inclusiveness of UK data and evidence.
- 2.1 ONS should establish a clear mechanism and timetable for monitoring and reviewing the recommendations of the Taskforce, reporting on how far they have been implemented and outlining strategies to ensure their implementation going forward.
- 2.2 Data producers should institute a continuing user forum to embed the work of the Taskforce into regular workstreams.
- 2.3 Data producers should engage with academics, user groups and others outside of government with experience of key equalities issues relevant to the UK on an ongoing basis, to share knowledge and approaches to measurement.
- 2.4 ONS should undertake a systematic review of how other National Statistics Offices undertake the collection, analysis and reporting of equalities data. They should work with other countries with promising practices to share knowledge and approaches to measurement and reporting.
- 2.5 Data producers should consider the joint financing of data collection across the UK data infrastructure, to ensure that the costs of addressing data gaps and under-representation are shared and sustainable, and that cost-effective solutions are developed.
- 2.6 ONS should play an active role in international initiatives to improve the inclusivity of statistics, including, but not limited to, the UN Committee on the Rights of Persons with Disabilities and the revision of the UNSD Guidelines and Principles for the Development of Disability Statistics. To take this forward, ONS should seek the establishment of a UN City Group on Inclusive Statistics.
- 2.7 ONS and other data producers should share, evaluate and publish effective innovative practices to enable wider learning.
3. Ensure that all groups are robustly captured across key areas of life in UK data and review practices regularly.
- 3.1 The Inter-Administration Committee and the UK Census and Population Statistics Strategic Group should set up a mechanism to regularly review who is under-represented in UK statistics or data collection exercises, and lead work to address this. This would enable them to respond to changes in coverage, alongside changing social composition, social categories and social understanding.
- 3.1.1 Particular current priorities are those who are digitally or linguistically excluded, disabled children and parents, victims of intimate partner and domestic violence, particularly older and migrant victims and minors, women experiencing pregnancy and maternity, children in food poverty, residents of communal establishments such as prisons, Immigration Removal Centres, hostels and care homes, the “hidden homeless”, and small groups such as Gypsy, Roma and Traveller communities, scheduled caste and tribe Asian and African groups who, at present, are largely invisible in published statistics.
- 3.2 Data producers should review the representativeness of key surveys and administrative datasets (initially by benchmarking against the UK 2021/22 Censuses) and take swift action to address identified issues, particularly as they relate to historically under-represented populations (for example young black men) or more marginalised groups (for example children), including but not restricted to those noted in recommendation 3.1.1.
- 3.2.1 Such a review should ensure that key surveys and datasets are using consistent measures (see also principle 5 on appropriateness and clarity over the concepts and principle 7 on harmonised standards), and that inconsistencies are not distorting comparability.
- 3.3 Data producers should explore how to improve the collection of administrative data on characteristics that are legally protected in equalities legislation in England, Wales and Scotland with users and relevant government departments. Such as, religion and belief, gender reassignment (gender identity), marriage and civil partnership, maternity and pregnancy, other relevant characteristics such as socio-economic background and migrant status. Additionally, regularly collected (and also legally protected in England, Wales and Scotland) characteristics such as sex, ethnic group and disability status should continue to be comprehensively and appropriately recorded.
- 3.4 Sex, age and ethnic group should be routinely collected and reported in all administrative data and in-service process data, including statistics collected within health and care settings and by police, courts and prisons. The quality of these data should be regularly reviewed to provide information that better reflects those in contact with these settings.
- 3.5 Where it is impractical or inappropriate to collect characteristics that are legally protected in England, Wales and Scotland in administrative data, or where such data provide insufficient information on groups and populations’ experiences, data producers should consider large-scale survey exercises to supplement understanding of these issues.
- 3.5.1 For example, supplementary data are likely needed to better capture information on sexual orientation, especially among those not of working age and outside of cohabiting couples. Additionally, data on psychological well-being across the age and sex spectrum are required, particularly on the mental health of older men, girls and young women.
- 3.6 Data producers should work in partnership to ensure that UK administrative data sources appropriately reflect relevant characteristics as much as possible (see also principle 5 on appropriateness and clarity over the concepts and principle 7 on harmonised standards). This includes working to link various administrative datasets, and to survey data where appropriate (see also principle 6 on methods used) to ensure coverage. This will enhance the potential of (linked) administrative data to fill the gaps for relevant characteristics, and their intersections across and within the different countries of the UK. This would also improve the understanding of relevant groups and their experiences over time and across settings.
- 3.7 Data producers should evaluate the coverage of non-private household population groups in UK data and take the necessary action to address those missing from the current data. In particular, ensuring longer-stay residents in care homes, hospitals, and prisons, and the turnover of people between private households and other (or no) residences is reflected.
- 3.8 Data producers should recognise the diverse data needs of different users in the collection of data about specific populations and groups and the intersection between population and group characteristics, and put mechanisms in place to ensure that data collection and reporting serves a variety of user and respondent needs.
4. Improve the UK data infrastructure to enable robust and reliable disaggregation and intersectional analysis across the full range of relevant groups and populations, and at differing levels of geography.
- 4.1 Data producers should ensure sufficient granularity of data to enable meaningful disaggregation. They should avoid the use of meta-categories which can disguise heterogeneity between groups, within them, and with which people may not identify (for example, White, BAME, LGBTQ+).
- 4.2 ONS and the Cabinet Office should actively promote an intersectional approach to exploring and presenting equalities data across the UK. Potentially misleading single characteristic analyses should be avoided, and “like for like” analyses controlling, for example, for age, sex, sexual orientation, racialisation, socio-economic background and position 1, and place be readily accessible.
- 4.3 ONS, National Records of Scotland (NRS) and Northern Ireland Statistics and Research Agency (NISRA) should carry out detailed intersectional analyses from the 2021/22 censuses, to provide granular insights into the nature of disadvantage. To include socio-economic background in the analysis of intersectionality, the analysis should also draw on other authoritative sources such as the Labour Force Survey/Annual Population Survey, since the important measure of socio-economic background was not included in the 2021/22 Censuses.
- 4.4 Producers of existing online tools should adopt intersectional approaches, enabling analysis of different characteristics together (such as ethnicity and religion) to improve understanding of inequalities.
- 4.5 Data producers should use targeted oversampling of under-represented groups as an approach to address specific gaps in knowledge that result from small sample sizes and to facilitate intersectional analyses.
- 4.6 ONS must ensure that the 2023 recommendations on the future social statistics system provides an enduring solution, to meet the full range of inclusivity data needs that are included in the recommendations, including for those groups identified as priorities in 3.1.1.
5. Ensure appropriateness and clarity over the concepts being measured across all data collected.
- 5.1 Data producers should review the conceptual foundations of their measures for relevant populations and groups, ensuring the measures that are used accurately reflect the current standards and legislation. Data providers should ensure that measures are conceptually robust and do not incorporate formulations that might be deemed to be derogatory, inappropriate or misleading.
- 5.1.1 As a priority, ONS should transition its measures of disability to approaches more firmly based upon the WHO ICF and ICF-CY biopsychosocial model conceptual frameworks.
- 5.1.2 ONS should transition its measures of ethnicity and religion so that they better correspond to the current conceptual understandings, reflect the diversity of the population and are recognisable and meaningful to those from specific ethnic and religious groups.
- 5.1.3 The robustness of measures to capture the experience of populations and groups should be considered. For example, to measure poverty more effectively, ONS should review income equivalisation methods, improve estimates of income poverty and fuel poverty amongst people with disabilities and other affected groups.
- 5.2 In cooperation with the Devolved Administrations, ONS should develop (and evaluate) a set of measures of socio-economic background that are suitable for collection in administrative datasets and surveys. At a minimum, this should include measures of parental occupation and parental education and be sufficiently granular to capture a range of occupational classes and educational levels, while not being burdensome for respondents.
- 5.3 Data producers should ensure that survey and question design is based on a clear conceptual understanding of the information that is required, drawing upon best practice to translate this conceptual understanding into accessible and appropriate data collection (see also recommendation 1.5 under principle 1 on trust and trustworthiness)
- 5.4 Data producers and analysts should ensure that the language used in the collection and reporting of all characteristics is clear. For example, clearly distinguishing between concepts such as sex, gender and gender identity; or ethnic identity and ethnic background. This would help to avoid ambiguity and confusion among respondents and data users, which can undermine data and analytical quality, as well as belief in the validity and reliability of data.
- 5.5 When sharing or reporting data, data producers should be transparent about how the data have been collected (for example, the questions, modes and mechanisms for providing responses, including clarity around the use of proxy responses). Comprehensive metadata should be published alongside their data, which are accessible to respondents and data users to enable them to assess the quality and suitability of data.
6. Broaden the range of methods that are routinely used and create new approaches to understanding experiences across the population of the UK.
- 6.1 Data producers should explore opportunities to utilise more varied, innovative and flexible approaches to data collection and combination, where this will be of particular value for enhancing our understanding of the experiences of relevant groups and populations and/or for enabling the inclusion or voices of groups currently under-represented or missing from existing data sources – such as undocumented migrants, those with disabilities, the “hidden homeless”, and children. Such approaches will also be relevant for providing more comprehensive information on the characteristics and experiences of those priority groups identified in points 3.1.1 and 3.2.
- 6.2 A wider range of methods should be considered for capturing those temporary experiences that are not often well recorded – but which may be important for inclusion. These include pregnancy, hospital stays, school exclusions, periods children spend looked after by the local authority, “sofa surfing”, and periods in prison or on remand. It also includes experiences that are sensitive and poorly covered for some groups, for example intimate partner violence and other forms of domestic violence, especially as experienced by older women.
- 6.3 Some of the currently underutilised methods that would provide valuable additional insight include: ethnographic methods to understand lived experiences, field experiments to understand more about discrimination, comparative studies across the UK to examine “what works” in promoting inclusive data collection, linking administrative data to survey data or other administrative sources, and better using and enhancing longitudinal and panel data collection.
- 6.3.1 For example, for those surveys currently collecting information about children, data producers should consider what information can be collected directly from children themselves, using appropriate instruments and diverse forms of data collection (for example, pictures and diaries), drawing upon best practice in data collection and ethical approaches, while recognising the potential additional time / burden involved and the privacy needed for children to be able to take part. Data should also be collected to reflect more marginalised children (for example, disabled children, children of prisoners, Gypsy, Roma and Traveller children, looked after children, refugees and unaccompanied migrants) as a priority.
- 6.4 In all innovations, such as those noted, adequate attention must be paid to issues of consent, trust and trustworthiness (see also principle 1 on trust) and risks of disclosure.
- 6.4.1 For example, in relation to data linkage, ensuring that the data to be linked, and the linking process, have been demonstrated to be of appropriate quality and accuracy and that safeguards have been put in place to protect respondent privacy, confidentiality and anonymity, in line with existing guidelines such as the UK Statistics Authority Office for Statistics Regulation’s Systemic Review Programme on Joining Up Data for Better Statistics. Open communication about the use of data and safeguards is also essential. Linkage and data security should follow best practice and be justified in relation to public interest concerns, while not using safeguards as a means to restrict information about or analysis of populations of interest (see also principle 3 on groups being robustly captured).
7. Harmonised standards for relevant groups and populations should be reviewed at least every five years and updated and expanded where necessary, in line with changing social norms and respondent and user needs.
- 7.1 Data producers should research user and respondent data requirements and draw on best practice standards and guidelines from other countries and relevant international bodies, to ensure that harmonised standards remain appropriate and relevant.
- 7.2 Data producers should undertake research into the user and respondent needs for data on groups of interest and provide guidance on how to collect this across different modes, continually reviewing approaches to maintain relevance. This information should then be used to update the existing harmonised standards.
- 7.3 ONS and others across government and the devolved nations should work together to improve the harmonisation and comparability of data sets across the UK, between regions and over time. They should ensure that the basic data are sufficiently granular in each part of the UK to avoid situations where the only harmonisation that is possible involves unsatisfactory “lowest common denominator” meta-categories such as “white” and “non-white”. This is particularly important since such binary categorisations are often experienced as pejorative in taking whiteness as the norm, rather than recognising diversity (see also recommendation 4.1 under principle 4 on UK data infrastructure).
- 7.4 Data producers should use harmonised standards when collecting data, or more granular systems which are compatible with the harmonised standards, to improve comparability and better use existing data.
8. Ensure UK data and evidence are equally accessible to all, while protecting the identity and confidentiality of those sharing their data.
- 8.1 ONS should work with others to create a centralised, explorable and accessible UK-wide “one-stop-shop” database of equalities data and analysis.
- 8.2 Data producers should make administrative data accessible to a wide range of users, including to non-experts. This should include both outputs and non-disclosive “raw” data to enable alternative analyses. In particular, data collected on residents of communal establishments, such as prisons and care homes, need to be made available in order to improve services and transparency.
- 8.3 Data producers should develop additional, user-friendly online tools for non-experts to explore existing datasets. Users should not be limited, as in some existing online tools, to pre-set tables provided in advance by the data provider. Rather, they should be able to explore the data so that it can meet their needs, subject to automatic disclosure safeguards, as for example, with Statistics Finland. On the principle of “generalised reciprocity”, where members of the public have provided their data voluntarily as respondents, data providers should not charge other members of the public needing access to the data.
- 8.4 Data producers should consider language, literacy, format and comprehension when presenting analysis and evidence, in line with the 2018 Accessibility Regulations, and produce accessible websites and outputs for diverse audiences, including the digitally excluded.
- 8.5 Where relevant administrative data exist that enhance the understanding of inclusion/exclusion, the responsible departments should be required to publish these. Data producers should, as far as possible, adopt an open data model, to help ensure that data are freely available and usable by everyone.