Guidance

Statistical principles for the Race Disparity Audit

Updated 21 April 2020

This document outlines the statistical principles followed in the production of the Race Disparity Audit.

1. Scope and coverage

The Audit will compile statistics about public services which can be disaggregated by ethnicity.

The Audit will compile statistics on ethnicity using information about people’s ethnicity.

Where data can be disaggregated by ethnicity, it will be included in the Audit regardless of whether it shows the presence or absence of disparities.

Data by ethnicity should always be further broken down by income and/or geography where available.

There will be tradeoffs to make between the granularity of ethnicity data in combination with income and geography data, but the primary focus should be on ethnicity data.

The Audit will capture information about a range of income and socio-economic groups broken down by ethnicity, but will not otherwise seek to define specific groups such as ‘the White Working Class’.

The Audit will present statistics at the smallest geographic level possible to help make it as relevant as possible to the public.

The Audit will present data on institutions and business units (eg schools, police forces, hospitals) to increase its relevance to the public and help drive service transformation.

The Audit will publish statistics on ethnic disparities for the whole of the UK, but comparability across devolved administrations will be limited due to differences in geography, public services and data systems.

2. Types of data

The Audit is not a repository for research on ethnic disparities.

The Audit will compile quantitative data on ethnic disparity from a range of sources including Census, surveys and administrative data.

The Audit will present descriptive statistics and may also refer to findings based on more advanced statistical analysis, such as regression, provided these are already published official or national statistics or peer-reviewed.

The Audit will collect and report time series, possibly phased subject to resources and demand.

The Audit will demonstrate commitment to transparency and build trust by publishing a forward programme for release of future phases of data and new datasets or analysis.

3. Responsible analysis and making meaningful comparisons

The Audit will disaggregate ethnicity data by income and by geography, and other key relevant factors (e.g age, gender etc) to expose differences and/or avoid misleading conclusions.

Departments will consider the granularity required to show differences which might be otherwise masked eg to analyse different types of health complaint or criminal offence.

The Audit will present data on the observed differences between ethnic groups without statistical adjustment, but departments may - on a case by case basis - perform the adjustments necessary to avoid misleading conclusions. While unadjusted statistics/rates will generally be used to show disparities, such as about processes, it may be necessary to adjust some outcomes for confounding factors eg age and sex in health outcomes.

The Unit will work with departments on a case by case basis to understand the complexities of analysing inequity in treatment by public services which depend on need and accessibility as well as other factors (eg personal choice and use of alternative services).

Departmental proposals for further analysis to aid understanding of ethnic disparities will be tabled in a public forward work plan.

The Unit will scrutinise departments’ statistics to ensure they demonstrate integrity, are meaningful, and highlight disparities where they do and don’t exist.

The Audit will communicate statistical analysis in a non-technical and standardised way, using the ONS style guide, specific good practice guidance (as advised) and digital platform content designers.

4. Commentary and presentation

Departments will provide supporting analytical commentary to include key headline findings and their statistical significance, and caveats around the data quality and analysis - being transparent about strengths and weaknesses.

Findings should focus on key statistically significant differences which highlight both existence of ethnic disparity in access to public services as well as parity of access.

The Audit advises use of the ONS style guide and specific good practice guidance (as advised) for consistently high standard analytical commentary.

5. Quality and quality assurance

The Audit will not seek to suppress data if it is poor quality but will instead provide necessary warnings and caveats around data if it is the best available.

The Audit will seek expertise from GSS Good Practice Team and Office for Statistics Regulation (OSR, the regulatory part of the UK Statistics Authority) on advice to departments on assessing and communicating quality.

The Audit will work with the Technical Working Group and other key stakeholders (including GSS Good Practice Team and OSR) to agree quality thresholds for inclusion of data in the Audit to meet its purposes.

Departmental analysts will assure the quality of their data, analysis and commentary, signed off by Heads of Profession (HoPs); and the Unit will seek ONS and OSR quality assurance for the overall product.

6. Identifying gaps and harmonisation

The Unit continues to work with departments and engages with users, academics and other key stakeholders to identify gaps in data and understanding, and prioritises areas for improvement.

The Audit acknowledges current inconsistencies in the classification of data on ethnicity, income and geography across public services and will make recommendations for harmonisation.

The Unit will ask departments to identify potential further data collection or analysis which could be done to help improve understanding and will make this knowledge publicly available as part of the Audit process.

7. Ethnicity

The Audit advises that ethnicity information is reported using 2011 Census 18+1 ethnic group classification where possible, but lack of comparability across all sources will be inevitable.

Pooling multi-year data when appropriate may increase the reliability of analysis by ethnicity, income, geography and other relevant sub groups but recognise there will be tradeoffs (eg time-series).

Departmental analysts should explore the trade-offs required in banding ethnicity, income and geography and seek input from the Unit.

As a minimum reporting standard, White British people should be distinguished where possible from all other ethnic minorities (including other White and White Gypsy Traveller).

8. Income

Income is reported using weekly income bands where possible.

Equivalised household income is used where possible to take account of household composition and enable fair comparisons.

Where there is no income data available, proxy measures may be used in the following order of preference:

1. Measures of individual or household economic well being (eg Free School Meals), followed by measures of socio-economic group.
2. Area-based measures of deprivation at neighbourhood (ie LSOA level).
3. Area-based measures of deprivation at higher geographic level, but not higher than a local authority.

The Indices of Deprivation are the recommended area-based measures of deprivation, and the Unit will work with departments to select the most suitable geography, index and summary measures.

9. The White working class

The Unit will not define ‘working class’ and will infer this using income data where available in preference to occupation-based measures (eg NS-SEC).

10. Geography

Geographical analyses should use statistical geographies, presented in line with ONS geography policy, or well-known administrative geographies.

Data should be disaggregated to the smallest level possible, and by region as a minimum, using multi-year data if appropriate.

The Audit will, where possible, collect data for Lower-layer Super Output Areas (not wards) and may use it to produce and report ward level estimates.

11. Business units or institution level data

The Unit recommends reporting information at ‘business unit or institution level’ (eg hospital, school, prison. The Unit will work with departments on case by case basis to understand the capacity and technical considerations which will be relevant to phasing of this data, and potential impacts.

12. Data formats

The Unit intends to collect aggregated, unrounded numbers and statistics and will work with departments to devise the most appropriate measures to highlight ethnic disparities which aid comparisons across services.

The Unit will develop processes for uploading data which will seek to minimise errors, reduce the burdens of uploading and checking data, and improve standardisation.

The Audit will include comprehensive metadata, drawing on relevant GSS guidance, and will seek to devise a sustainable system for collecting standardised information from departments.

13. Disclosure control

The Audit will not hold data which is identifiable or disclosive and will not hold any records on individuals even when these are anonymised.

Departmental statisticians will ensure there is no disclosure of identifiable data, signed off by the Head of Profession, and will inform users about disclosure controls applied to the data.

14. Future updates

The Unit will work with departments to better understand how to prepare for future updates, including their capacity for APIs.